Bringing the power of Google Search to every business (Cloud Next '19)
For. Those of you who are not familiar with the product, we announced the product, at, the last cloud next in San Francisco. Late, last year and the, product has been generally available, since. November, of last year. The. Focus of today's session, will. Be to understand, the power of this product as, experienced. By our, valued. Customers and. Explained. By our. Partners. That we are extremely lucky to have and have. Here with us on stage today, so. Today's presentation will, in fact be led. By some of the best partners we have in North America. Talking. To us about customer. Deployments, in, globally. Recognized, large, corporations. So. We have a really simple agenda, I will. Walk you through our cloud searches in a nutshell. Hand. It over to partners one-by-one will walk us through a, series. Of deployments, that they've had walk. Us through the. Experience, with the product, with the API so on and so forth from that perspective and. We'll. End with a few exciting, product announcements, again from me, so. What is Google Cloud Search. In. A nutshell we have tried to bring the power of Google to every business, imagine. Being able to index. Everything. In a company I literally. Mean it index, everything, in a company. That. Is what will ultimately allow you to bring the information that, your employees, in a company need as. And. When they need it it puts information at their fingertips to. Do that, we, have to put privacy. And security at the core of this offering, so. A lot, of innovative. Engineering has gone in to reflect. Permissions. From the source in mere real-time into, the index so, that users, never see content that they are not supposed to. Obviously. The whole product is powered through a lot of machine learning from ingestion through ranking, to providing, answers so on and so forth the, goal here, is to bring every, aspect, of the convenience. Scale and speed of, google.com. To every business. Feeling. Done in a little bit this, is what is there in the product but, the core of it all is a permission, or an actual sensitive, index this, is not a regular content index this. Maintains really rich information, it maintains interaction, data it maintains, lots. Of extracted, insights from content, signals, in terms of who you are as a user so. On and so forth to rank documents. Uniquely. The way Google can do to. The left of this is an API we call it the indexing, API and, an, SDK, that we provide for, indexing content from any source on Prem, in the cloud or sassed applications. If, you happen, to be a G suite customer. You. Can index. This content automatically, we don't you don't need to do any work we just do it for you automatically. Obviously. The product is available as a standalone offering which means you do not need to be a G suite customer to use this product on the. Right you, have a query API that, allows you to implement search applications, in the context, of the, workflow surfaces, that your employees use. Core. Benefits. Starting. With the most obvious we. Bring world class search out-of-the-box, ok, close. To it that's, something that we've learned through our initial, deployments, and pitches is this. Product is truly serverless, no, concept of nodes no, concept of buying more storage no concept of figuring out how many VMs you need to spin up this was truly survivalist, just keep indexing, what you want keep filing up applications, it just, works.
Hardened. Security, that have touched upon we. Bring, truly. Enterprise wide search we really invite companies to index all of the content, right and of, course a product is internationalized. Works 400 plus languages out of the box and it's compliant with standards, that most enterprises, care about I. Would. Like to emphasize that the product is available as a, standalone offering, any. Customer. Any business in the world we invite you to try us as a product, that's. If you are a G suite customer we give you the added benefit, of indexing. Your G suite content, without any effort, but, even otherwise anybody, is free and welcome to use it, the product is priced on the basis of number of documents, number of queries and, we. Start. Off with packages, of various sizes and then, if you go beyond those quotas we have a very very simple, usage based model that kicks in after that. With. That background it's, now my privilege to invite, our. Esteemed. Guests today on stage Chad. Johnson, from sadha Brian. McKay from onyx, Eric. A moment from Perficient and cameron contra max incher please come on stage let's give them a round of applause please. So. Just to set up. These. Gentlemen for you know what they're going to be covering with us today, just. To give you a quick sense each of them is going to be covering a different aspect, of the, product. As they, have experienced, it their talk will touch upon many different aspects but I've just called out some, core aspects, that for each of them. Onyx. Will be talking to us Brian, from onyx will be come to us about the. Rapid, deployment experience they've had in the product. Eric. From Perficient, will be covering such the levels out of the box. Charge. From sada we'll be covering the benefits. Of having an effortless infrastructure. And getting. Good search quality out of the box and finally. Cameron from Accenture will, be providing a very interesting perspective on how, search can be used as a building block for larger, and more complex content analytics solutions, with. That overview I invite, Brian, to take it over from here. Giants. Thank you good. Afternoon thanks. For joining us. Hi, everybody my name is Brian McKay I lead our cloud search practice, within onyx we, were our, company was formed in 1992. And, in 2002. We. Became one. Of the first Google, Enterprise partners. So we've been in the search business for a long time since. Then we've grown significantly, both, in terms of products whoops. Products. Both in terms of products, and services, as. Well as a company so, we have over 200 employees now, we're, serving over at the housing customers, we.
Represent. And we resell, all the different products that Google has to offer the whole product line so we're not just search, anymore, although for, me in my group it is the thing that we do and. The. Other thing I should mention is we are a three-time. Recipient. Of the. North. American Partner of the year. Little. Outer onyx folks. So. I've got two use cases that I'm going to talk about this, afternoon. The first one is Philips, Philips Healthcare and, so. This is, a this is a very common, use case that. You'll find in a lot of organizations some, of your organizations in fact may have this use case this, is a field, used field, service use case it's. Really filled, service knowledge management and, customer, support, and in, 2015. They came to us with a problem it was primarily knowledge, management and, customer, support and the, challenge that they had was. That for their service engineers, it just took far too long than to find anything, in, order to be able to service their customers, in. The field so within. Philips this particular group is about 5000, field service engineers. All. Around the world and their, job is to go into hospitals, and look after the, MRI machines CT, scanners, ultrasounds, etc, etc so, they've. Got a pretty important role out there in the field their, challenge was they, were spending evenings and weekends, trying, to find the information they, needed so that they could service the equipment, the, day before they, were going in, actually work with the customer work with the equipment it, was just taking too long they couldn't find what they needed sometimes, they found that sometimes they didn't all that adds up to equipment. Downtime extended. Downtime, and. Just in general the inability to get the machines back up and running and that was a that was impacting, some really critical metrics, for the customer, you think customer satisfaction. First-time, resolution. The time it takes to solve a problem etc etc so. Working. With their team we. Solved that problem for them and we. Were able to help, their team realize, some, really important benefits for example. Reducing. The amount of time it takes to search for information and, find what they needed by, 27 percent and the other one which was I think even more important, and important, for their customers, reducing. The time is solved by 15% so the time in the field it's much more valuable they got the equipment back up and running up time increases, etc, etc a, very. Important milestone for them in. 2018. They came to us with a different challenge this time the, challenge was they had to move off of an old solution, and migrate over to a new solution. And. They had a very compressed, timeline within, which to do it so, this was a migration exercise, primarily. But. They had some really critical objectives, so number one they, wanted. To ensure, that they had a minimal, footprint. In the data center, essentially. They are moving to the cloud and. Not. Only were they moving to the cloud the other issue they had was, just the cost of implementing, within the data center and the.
Amount, Of time and effort it takes to implement it with from the data center so, challenge there is we, had a very so. We had to find a solution that allowed them to move to the cloud the, other one was they, needed to be ready for you 18 3 months this is the rapid implementation, part of it and, and. That was a very significant, challenge. But. I've. Had Shawn some of the details it was very successful the, last one is you. Know since, they were migrating, from a solution they wanted it to be transparent, no training, no retraining. And they wanted the solution, to give them a search experience that was equal to or better what they had with the previous solution so. First of all the man is serviced I can't. Underestimate this. You know that notion of a managed service there, are no servers to setup there's, no software to install as, giant had mentioned there's no nodes you're, not tuning, and tweaking the, search engine it's just there it just works so that's, the first key point and the second one is with a robust. Connector. Framework and api's, we. Were able to actually start development remotely, and help. Them move along a lot, faster and get the, get, the deployment, ready within that three-month window and into, their their QA environment. The, other thing we should mention is the performance of the connectors, you know indexing, in days not weeks is, a really, important measurement and the, last is once. The system went live, nobody. Really noticed and. That was a good thing because that meant we hit the objective, that meant that we were giving them what, they had if, not better. Now. When you take a look at the architecture I, do want to highlight a couple of things here the. First point is, years. A few years ago when we started this almost, every, content source you see on this diagram that's on the left was. On premise over. Time they were moving everything into the cloud the. Other thing is, everything. On the right was, also on premise, so with this implementation, what, we've done is now we didn't obviously move their content sources to the cloud but, what we've done essentially. Is move the entire solution ahead of their data center and into the cloud the, only thing that we that's, left in, the data center that we put in their data center are, those green, boxes, and. Those are the connectors those are the connectors themselves, running, on virtual machines so. That's really all that's left I mean that's there for security reasons and you know so that it's close to the on-premise sources, but, that's all it's left everything else has been moved to the cloud that greatly simplifies. Solutions. And allows, us to move a lot faster. Now, this is a view of the user interface looks. Like a fairly typical enterprise, search use case but there is something here that I, want to highlight and it's really, important and that is on the left hand side so everything, on the right looks like fairly, typical you. Know Enterprise, Search stuff but on the left, the thing that's really important, are those filters, within. Their solution. Almost. All of their content, did not have those filters so as we were ingesting, the content, we had to add those, filters we, did that in the connectors, in their. Previous solution, we. Had another. System, which was a third party offering, they, analyzed the content, and enriched the metadata with these connectors, and the processes, that we created we were able to add that metadata with the connectors, and that's really important, because that's, how their engineers, are able to find their information so, quickly oh. One. More thing I wanted to mention and that is, if, I just go back I. Said. There. Are some best practices, around, metadata. Enrichment. And, that's. Not the stuff I'm gonna dig into deeply, and I know that Eric is so I just wanted to make sure I'd feed that up for his conversation. The. Second use case I want to talk about is whirlpool for those of you who joined us last year on stage or sorry who joins last year for the for the launch of cloud search you've, heard about this use case the.
Thing That's really important, about this use case is, really. It's. The why it matters, right in the engineering world they tend, to have systems, very diverse systems, in the case of whirlpool there's 21, different systems that. They need to search people have to find information and all these different systems that takes a long time but, the key here is that with a solution, like this they have a single pane of glass to. Search across all those systems and so, finding, and reusing, knowledge, they've invested, in is now possible that's, one of the key benefits the, other of course is you start to get into things like the, reduction, of design. Or the reduction of cost of design in an improvement in time market. Now a rapid, implementation, as. Giant. That highlighted is really important, and is, another key, part to it that's really important, and that is how quickly, now we can index content, now. We're seeing rates of. About. Twelve million, documents I, think I think, the API is are capable of about twelve million documents you know in a week at, whirlpool we indexed their documents, twelve million documents in, just over a week and that's, super. Critical because. As I'm about to show, you here, there are some things we're working on next, right, in, terms of significantly, improving the experience. For their users which. Is really critical because everybody. Wants that Google experience, well now we're about to take that step but, it's the performance, of the connectors, that is an enabler, if you can index your content fast. That, means that you're able to experiment you're, able to go and start to experiment. With you know the metadata and the filters the facets, the operators, and everything else we know we didn't get that right for whirlpool their users told us not so, we have work to do and this. Is what allows us to get that right for the users the other thing that we're. Starting to do is to. Introduce. A. Really. Really, exciting feature probably, the probably. One of the ones that I personally I find is the most exciting, and that is the the ability to. To. Start asking your, query, as a question, query, interpretation. Or query, intent and the other one is there's a lot of documents, out there in an index that have knowledge, buried. Deep so, we want to be able to start surfacing, that knowledge, as an answer to a question those, are the things that we're really looking forward, to. So. I just want to wrap up by letting you know there's, a few of us who are here in the audience and. Afterwards. We're gonna be here if you have any questions about the work that we do our, company, or the use cases please come see us afterwards. Thank. You. Thank. You very much Brian for that so very quickly they'll be now switching over to proficient. Eric. Will speak to us talk about a couple of their deployments, and help. Us understand how the search elements out of the box in cloud search help their customers thank you very much. Thanks. Giant so my name is Erica Merman I'm our director, of our search and content practice, and proficient high, level proficient, is a North, American focused. Services. Integrator and, software, integrator who has about 3000 consultants in North America our search practice, has been around for about. A decade now we've done over 200, successful. Implementations, and have around 30 search, focused. Clients. Who are consultants. Who are really focusing on search and I in particular, so. What I want to talk to you about today is a couple of different customers that we've worked with and and how, we've used Google, cloud search to kind of improve, relevance, and give them a great experience out of the box all right so the first one I want to talk about is a large, financial technology. Services customer about 22,000, users of this system so, as, you can see from this slide they're looking at a broad range of content right so eight different content sources crawling, websites loading, from ServiceNow, or Salesforce, crawling.
Jive Or confluence. Right all, of these are kind of heterogeneous, data sources that don't all have the same metadata, and information. For them and we want to give them all into a single pane of glass right. That single pane of glass is hosted on a SharePoint environment, it's, got seamless, single sign-on with Google to automatically, log you in know your preferences know your history and so on and it, really operates, as a unified, portal for all of that content all right so that's experience, number one experience. Number two is another customer we're working with who are, 350,000. Users so very, large food and beverage, manufacturer, right and as, part of this we've synchronized over all of their groups and their permissions, and their users right where we, have the ability to to keep permissions, on every one of these documents across thousands. And thousands and thousands of groups and then, they're loading lots and lots of content in so 21 different content sources for phase one which, is just the tip of the iceberg in this case things. Like databases where they're loading out of their own custom applications, that they've written internally, things, like I'd be IBM sweet so FileNet and, Wickham and portal or Documentum. Or online SAS services, like get abstract, or ebooks 24/7, that they use for learning they, are taking, these into a bunch of different application, experiences, where each has a different sort of way of working so, basic. Level this is something like people search being able to find what. People exist in my organization, and how they relate to one another right next. Would be something like research, library, helping to helping. Their R&D folks understand, what R&D they've done in the past and use that to build on top of the, R&D, they're going to do in the future unified. Enterprise Search once again we're looking to combine all of these sources together and really, use this as a way to accelerate. The, the time to value of both the product and also to allow, their users to very quickly find what they're looking for, now. In both of these environments, as I've mentioned these are kind of heterogeneous, environments, where there is a lot. Of different content that always ranked relatively differently right so, when, you run into this scenario a lot of times you, know the base way is you just say alright I'm just gonna look at the content, you know contents, equivalent across all these let's search my content, but that doesn't always provide the best experience as. Giants mentioned before there's lots of signals that can be used by these engines to, help, influence, relevance, the context, behind that content so something like when was a document created or is, there a good description of the content or does it have tags or labels or, keywords all, of these things are pieces, that can be provided, into the picture to give a better, relevance, out of the box than just simply looking at content and understanding what's there, in. A lot of search platforms, so a lot of the open source ones or you know other competitive platforms, the way that you do this and the way you integrate, this content, is via, the query, end of the matter right essentially. They're what's known as schema list systems so I throw all of my data into the system it figures, out the types of fields everything is and then when, I'm querying what I'm doing that interface, I am essentially, tuning, and tweaking and saying okay for content source a title is very important, content source B description, is important content source C maybe nothing is and this is going to happen in every organization because, you might have a website, or a system that's run by marketing, and there's been SEO, out the wazoo done on it and you know you really have very good content whereas something else might be user created content where the only reason they're putting a description is because it's required for them to upload the content on the first right, you get very different strength signals, out of those two different scenarios so in this data in this model, when, you're adding additional, data sources you're essentially getting an almost an exponential, sort of problem right the more applications, I have the more data sources I have the, more complex that logic has to be the more I have to work, to really understand, all the different pieces of data that I'm ingesting with.
Google Cloud Search however we do it a little bit differently so we design, the, relevance upfront by designing, a schema and specifying, a what, type of data is in the system right so my field might be a string and integer a date and so on and that's the simple stuff then. We go and we can do and say this is how important, that field is so at a field by field level, within the data source we can provide relevancy, context, that Google's algorithms can use to say oh in, this data source I really care about the description, or the title or maybe the keywords or the tags right. After, that we can then provide operators, and operators are essentially, a shortcut, for how do users, know this field right so that Google's natural language and can take that and really. Understand, ok the when the user says they're searching for a document, with a description of blank okay let me rewrite that query to understand, here's the type and description, is one of the fields so let's rewrite the query to make it more accurate for the user even though they didn't know how to write, that query themselves, okay this, all takes the onus of driving, relevancy out of the query and puts it at the datasource end which is a honestly. A better area to do it simply because a lot of times the people who know their data best are the owners of those data sources so, we're, looking to kind of up front define, this that I can go and combine later and I, don't have to go and kind of retune, everything, after the fact I just have great out-of-the-box relevancy. So. Beyond, that now we have the case where there's multiple applications, I may want to load right so I might, have an engineering search I might have a Enterprise Search engineering. Search maybe I care a lot more about github or bitbucket or, JIRA right, the. General Enterprise Search I might care about my policies, or you, know what's the cafeteria, menu in, both of those I can have the same data sources index and use both of those in both searches but different importances, are going to be there based, on which search the user is coming from google, cloud search allows us to out-of-the-box, via just a toggle turn on whether a data source is a pair available for an application and then, B we can kinda go and set the importance, for that data source as part, of the application itself so. For engineering search maybe I don't care about policies, as much as I care about source code for, Enterprise Search maybe I don't care about source code as much as I care about policies, very, simple little tweaks we can do in order to tweak the relevance out after, the fact without having to go and change my queries and change my logic and so on combined, with the machine learning and personalization this. Really gives us the ability to have a great, out-of-the-box relevance, and you, know really, great, experience, and quick time to value for users getting a beneficial.
Product Into their environment, ok so, I appreciate everybody's time today if, anybody wants to come learn more about this or talk with me of my, contact, information on the screen my colleague Joe is also here if you'd like to talk and it's been great talking with you thank. You very much. Thank. You very much, Eric for that overview, now. Over to Chad from SATA to. Talk to us about a very interesting use case where the product has been used for an external search problem. All. Right thanks, everybody I am, the director, of enterprise search at sada. Sada. Was founded about 19 years ago and we. Were one of Google's original, launch partners, with their, enterprise partner program. At, this point we have been a multi-year, recipient, of their partner of the Year awards, we cover all of the, Google Cloud, suite, all of the product lines. Within. Sada, I run our search, practice, we have a team that, implements. Projects, just like this we, help customers. Figure. Out how, Google. Cloud Search can, meet their needs and then. We take, the. Project all the way through to, implementation. We. Take. A, very. Custom, approach to each project and I really have. Been doing this for about ten years now with Enterprise Search and really, enjoy all of the different industries that we get to bring search into there's a lot of things, that are common, across all the projects, but, each one has something it, has a story it has something that's interesting and I, want to tell you about one of those projects, today and this was the first time that I've worked in this a. Slice. Of. This. Type of search project. We, partnered, with a company called Avenue, insights and analytics and. Avenue. Is a public. Sector focused, solution, provider and they, developed, an application called, super, index super. Index is. Essentially. A search on top of real property records, for counties across the United States in particular, for, this project we were working with Oakland, County Michigan. Super. Index is interesting. Because we've. Probably all had this this, problem we've we've tried to find something on our local governments websites, and maybe, they restrict, what we can search maybe, they restrict how, we can search when we can search Oakland. County has made this very accessible. To the residents, of their county. Documents. Are. Scanned. And run. Through an optical character, recognition process. So that the entire mortgage. Or deed or, lien is searchable. Metadata. Is extracted, from those documents, and. Entered. Through. The the, registrar's office, and. At. This point the the the. Data comes out in. A structured. Format it. Was. I'm. Even sure what it was stored in to be honest but it came into a. Connector. That we developed, using Google's, software. Development, kit for cloud, search so. Within, just a couple of weeks we were able to take in massive, amounts of data from, these feeds coming, in with the full-text of all of the the, mortgage documents, all of the deeds all the liens. The. Metadata, comes, in there were over 75. Different properties. And attributes for, each of these grantor. Names subdivision. Names the, year that the the property, was built you can imagine there was a you. Know I had to have a large. Spreadsheet, just to keep track of all the different attributes. And fields for all these documents, we, designed a schema just like Eric talked about to. Map. All of these two different data types to have. Different importance, levels, for the different fields. All. Of this was fed up into, Google cloud in a secure index. Security. Is obviously not super, important for a public site search but this is a solution, that Avenue deploys to many different counties so, we don't want one County to accidentally, be searching somebody else's records so there is still security. And silos, even. In a public search, on. The right hand side you'll see that the output is. Basically. A search results. Page I'll show you a larger, picture of this in just a second but there's also an interesting alternative use. Of the search on the the output end. We. Maybe, not have all had this happened to us but it can happen where, property. Records might be filed, maliciously. Maybe, someone tries to steal. A property, or tries to. You. Know do something with with bad intent. Oakland. County and the super index allows the, residents, of the county to register. For. Alerts. That let them know if. New documents, are added to the system with, something. In their query they can be fairly, arbitrary, in, the way they specify, this they. Don't have to put just their name maybe, their name is you, know John, Smith and that's too broad so maybe they want to be able to search by their address. Or by the. Subdivision. To narrow, the search if, any new documents, are added they get an alert every night the system runs, those searches, automatically, you register, with your email address and you'll get a nice email letting you know if anything new was added it's a great proactive.
System, To, prevent fraud. And at least detect it early so let me show you a little bit more about the interface this is a quite. A complicated, search screen. There's a lot of different ways. That the user can filter the search we, have the facets, down the left-hand side like. Brian talked about we have the. Different sorting. And filtering. Options. Within. The search results, themselves, in the center there's. A, preview, of the document so they can see the. Page of the document they, can then purchase. And order full. Copies of the document, through, an additional process but, at least they can tell if they found the right document, there's a lot of metadata that's shown for each search result and it's cloudsearch that's returning all of that metadata, we, put in those 75 fields. We, indicate, which ones we want to come back in the search results, there's no limit we can have as many of them come back as we want the. Application renders, them in, various, panels and you know there's things that expand, and contract and. Let. The user verify. That they have found the right document, before they purchase it this, is a public site as giant mentioned, it is at ocmi, deeds comm, which we've included on the slides if anyone wants to check. It out and it, is publicly accessible. The. Second interface, that I talked about was the alerts this, is a pretty, simple interface, where the user types in a couple of key, words at the top. The. System. Automatically. Runs the search every night keeping. Track of any new results, so it's noticing, that maybe yesterday. That search returned four results, today, it returned five it, sends, an email to let us know that a new, document, has been added and if, it's something you didn't authorize. You can do something about it so. The. Biggest, reason. That, Avenue. Implemented. Google cloud search was that they are not search, engineers, they are not search experts, but, they wanted Google quality, results, without a lot of work they. Get that with Google cloud search they, know that their users will be familiar with the system it will work like Google, it's become that verb that we all use and that's what they wanted they didn't want to build their own they didn't want to design their own they, just wanted to implement a service, that provided very good quality results. Right out of the box in. Terms. Of infrastructure, they were having some pain points there was a lot of infrastructure, to manage their own search engine they. Were having interruptions, and down times during, certain maintenance changes. They, would load a new schema, they would push, a large batch of day and they would occasionally have downtime, during, those changes. Google, Cloud Search has eliminated those, for them and finally. The the schema, did. Support, the, robustness, that they needed it had like I said quite a few attributes a lot of different filtering. And sorting and faceting, capabilities, we were able to accommodate all of that for them in. Terms, of performance. And achievements, we did index, over. 13 million records, it took approximately two. Days these, were not very large records, they weren't you, know hundred page PDFs, but they were, you. Know there was a lot of data there was a lot of metadata the, full OCR, text of every document and we. Were able to run I want. To say that we were more the bottleneck, than cloudsearch was we parallelized.
The Indexing, meaning we we, had multiple, instances, running in parallel and we. Were not hitting, the. Quotas, that Google imposed we. Were way under the thresholds, so, I. Mopped. That, if we ever need to I'm optimistic that if we ever need to index even more than this that, with, the right infrastructure, and the right planning, on our part I I think we're always going to be the bottleneck not Google in terms of indexing, performance. We, are still getting approximately, one second, per query which, is a great execution time it's fast enough that users. Don't notice it. Responds. You, know feels nearly instantaneous. And. In terms of an implementation. My team was, able to achieve this in about eight weeks start, to finish which i think is fabulous it was, as. Eric said sort of a rip and replace it, was swapping. Out the old for the new but. That's what we do a lot and we just want to make sure that it's as transparent. As and as seamless as it can be for the customer. Lastly. I just wanted to give a little first-person. Quote here from Lisa. Braun who is Oakland, County's Register, of Deeds and Clerk and, he, gives a little bit of the history of why this is important why this used. To be difficult why, it was a struggle, even going back into the 1800s, in the way that records. Used to be recorded, and, Ledger's, and could. Only be looked up certain. Ways so from. Lisa traditionally. Online, land records databases, are limited, to searching pre indexed, fields, like, grantor, and grantee, now. With, super index oakland, county's google cloud search based solution, users. Have the flexibility, to search our records by any term and in, one of the five languages, that we offer this. Enables us to accommodate, our diverse population. Super. Index is user friendly and in the google format, that people are familiar with i just wanted to give a shout out lisa has joined us in the audience today i want to thank you for traveling, in if you have any questions about this she's very passionate about the. Benefits of this solution and we, look forward to more partnerships, with avenue. And with oakland county so thank you. Thank. You so much ad as, the last speaker but. Not the least important, it's Accenture. Camera. We'll be covering a very interesting set. Of deployments. But also walking us through how. Accenture, sees search as a building block into, a much larger, content. Intelligence, kind of a play in in a lot of emerging use cases with your camera. Thank. You Jonathan thank, you everybody for attending this, session really. Appreciate, the, chance to speak to you today this. Giant mentioned, I'm going to talk about three things principally. The first is, quick. Background, on, where. My group the search and content analytics group, which is effectively, the search practice, inside of Accenture, where, we sit and how we interface, with the rest of the organization, there I want, to talk a little bit about some, of the assets, that we built to. Help deploy. Search, solutions. And particularly cloud, search one. Of those is in an, interesting area, that I think is going to be more, prevalent in the future and then I'll really quickly. Touch, on three use cases that we've implemented over the years. First. Off I have, been Accenture for. 19. Months now and I came to Accenture, by way of an acquisition, I was, the founder, and CEO of search technologies, which was a small. Ish. Search. Consulting, firm that grew to about 200, people and, we were required by Accenture, about 18. Months ago and so, now we're four hundred and sixty thousand, people as an. Interesting. Exercise. So. Today we're, still focused, on search but, we're also searching, on. Content, analytics which is really what Google calls document. Understanding. And I'll get into that in a little bit more detail, but we've been working with Google, for many many years as search technologies, and now as Accenture.
We. Have many engineers, not. Unlike these guys who are trained in all the aspects, of Google cloud search, the. The thing that I like and the reason that we were acquired by Accenture, is that, the projects, that we're doing are, really smaller. Components. Of very large projects, that use. Many. Other assets, that Google has today so search, typically, it's part of a much larger implementation. So we have access to to, all of those groups and all of those customers, now through Accenture, um. Let. Me talk a bit about assets. When. We were at search technologies, we found that implementing. Search really required, us to build, accelerators. And one, of those key accelerators. Was in the area of content, ingestion, I think everybody most. People here have heard about connectors. So, we built a framework that we call a spire that has 40, or so connectors, that in some, cases overlap, with the Google connectors, but in some cases are complimentary because there are so many different. Document, sources out there so we have all of these connectors, available and I'll talk about some. Customers. That are using them and we. Are maintaining these and building, some great solutions with, some or in some cases all of these at Accenture we have 50. Content, sources, that's powering the intranet for four. Hundred and sixty thousand people so that's the first set, of assets I wanted to talk about now. This is a bit, of an eye chart and don't worry I'm not going to go through every box here but what I wanted to talk about was this area of document, understanding, if, I look at the project, that we're involved with today at Sencha, the, vast majority, are search. Huge cases the ones that we've heard about once. I'm going to talk about but. There's a there's a new growing, use case that that, we call content, analytics, or, document. Understanding. And this. Is where, rather. Than focusing, on search we're using natural. Language processing to. Understand. Documents. Once, we can understand, documents, which is effectively, tagging, them so that we've got much, richer, metadata, we can do many many things with with that content, we, can put it into a knowledge graph we. Can index it with search which is what I think most people are going to do but, we can start to build solutions, that are not search like such, a match is a good example we have a customer where we're matching jobs. To, resumes, automatically. By understanding, those documents, we've got a customer, we're working on right now where. We're routing mail, that's, coming in thirty five million pieces of mail we understand, that mail and we send it to the right Department, or the right person, so, document. Understanding, I think is going to be a very, very large area, with, that in mind google has launched the google, natural, language api which. In my opinion goes, hand-in-hand, with search, the. Other asset, which is in the yellow box that we've implemented or, we've built and we've just released it's a product called saga and and rather than go into detail about it I'll just say that it's a business, users, interface to help you build NLP. Applications. So if you want to build an NLP application, today you're gonna use Python you're, gonna have data scientists. So we're gonna build that algorithm. Using. Some of the algorithms from Google, and you're, gonna implement that as a model saga, enables, business, users, to use a drag-and-drop interface, and effectively. Have the code written for you so it's it's meant to democratize. Access to. This document understanding. API from Google which, we can then blend with Google, Cloud Search so so my own point of view is that search. And natural, language processing have, always gone together and I think that they will come together more and more and we'll see much more powerful, use cases as time, goes on. Let. Me really quickly touch, on three different case studies so this is a, motive client that we've been working with and the interesting, thing about this is it's a dealer extranet, so rather than the intranet, which is also powered, by Google. Cloud Search they wanted to build a brand new extranet, for thousands. Of their, dealers, and so, this this, was a big, multi-million. Dollar digital, transformation, project, so they're migrating, the, CRM, which was adobe experience manager, from, on-premise into, the cloud and my, colleagues at Accenture, are doing that and we, fit exactly in. The middle so search is of critical component. To this extranet, so we've been working, with this larger, team hand-in-hand, to implement, the Google, Cloud search solution, we. Had to use some, of our connectors, some of Google's connectors, but we also had to build a new connector, for adobe experience manager in, the cloud so the, the.
Integration, Of our connector, with Google Cloud search was particularly, you. Know particularly interesting, part of the project but I would say one, thing I had to pick out of this, project it's the integration with security. So, this, particular customer is using the IBM Tivoli. Product. And we, were able to integrate. Google Cloud search with that and that was a key, component of this so I would say that, the security, aspects, of Google Cloud so it's really important. In this this particular area, the, second customer, video-gaming. Customer, I'm guessing that you probably know who these customer I've set the first one I'll give you a clue they're on two wheels not for the, the second. One is a video gaming company, an. Enterprise, search application, your traditional, enterprise search but. Many different content sources including, loom apps that. They're using for their CRM, SharePoint. Both on-premise and online and again we were using the connectors, but the key thing about this was relevancy, they, really, needed strict, relevancy, and personalization. So, a marketing. Person in Austin, needs, to be served up with content that's more relevant to them from the content sources than than other, documents. And so that was a key, component and the relevancy, model within, cloud search really allowed us to do that very well again aspire was used here as a as, a an, asset to help us accelerate the, development here, we went live and this is going, to be I think a very large implementation. And, is finally a large. Aerospace client. Commercial. There's, only two of them so you have a 50/50, chance. I'll. Give you a better clue they're not in the news for the wrong reason at the moment. So. This is this is a big project, this this company, has moved from. Microsoft. Over to the entire g3, and it's a massive, transformation, for this organization and so, we're now implementing, Google Cloud Search we're at the concept, stage because, the G suite migration, is taking, a lot of time and we're, doing all kinds, of different proofs. Of concepts, and the one that I'd say that it's really interesting is we're using the people search capability. And that's, working really well but if what we're going to do is. We're going to start using that same capability but. We're going to use it for parts so that we can do some entity extraction and this is a perfect. Example of, where we're, going to start with search but, we're going to layer the Google. API the NLP, API, on top because. The document, understanding. Requirement. That this customer, is is enormous, and when we connect that together with search it's going to be a very very powerful system they, use a knowledge graph today so they want to get the Google, style answer, when they put questions, in to the intranet. Across, the whole of G suite. So, that's that's really the three things I wanted to say I just you, know I close. By saying that for, me I think Google Cloud Search is a very, powerful product, and when, you combine that with the Google NLP. API I think, you're going to be. Well. I'm already surprised, I think I'm going to be more surprised, about the kinds of things that we'll be able to do with those two products together so. Thank you and we'll. Be here at the booth if anybody has any questions. Well. Thank you very much Cameron for that and thank you all the panelists just summarize we heard use, cases of rapid, deployment with, Philips. And whirlpool amongst. All the benefits that, Brian. From onyx walked us through. Eric, thanks for that great description, of how such relevancy was really useful and how you could improve, search relevance by playing around with the structured, data fields. External. Intranet, effortless, infrastructure. Speed, of ingestion all of that thank you so much Chad and thanks. For the really nice presentation, on how NLP. And cloud search could come together. We. Were the last section of today's presentation. Because, I tend to speak better than I stand I'll stand. So. I have some really exciting announcements. About the product to make as. We close. The session today first. Moving into general availability starting. Today a whole, new set of machine. Learning based models, are beginning, to kick, so, what sorts of impact are we seeing we've been doing lots of experiments on these models internally. And we are really pleased to say that we. Are seeing a greater than 20 percent improvement in, the, mean reciprocal, score for, those of you are not familiar with that metric it roughly measures how frequently, do you does. A user see the results within the top three top five positions like units it's a measure of how, correctly do we pick the results had mattered the most fifteen. Percent left in successful, searches we know successful searches when people have actually clicked, on a link and are spending time on the link so we know that the search was successful, 15 percent lift, almost.
20% Drop, in abandoned, sessions, and. Along. With that and almost 30%, reduction in queries, where users do not find any results because, we're a relevant ranked engine we of course think that noisy. Results are just as bad as, poorly. Ranked results so we actually don't show results when, we are not sure the, way they increased, the, coverage, is by. Getting more intelligent, and aggressive, with regards to retrieval, of content, by applying a lot of machine, learning when it comes to knowledge understanding language, understanding. Content. Understanding and being able to retrieve results more aggressively, than otherwise so. We're really pleased about these results these search. Quality experiments, take a lot of discipline. So we've been experimenting, these for a while and we, hope that these, improvements customers. Will see an even, better experience, out, of the box what's with cloudsearch. Ii, there. Was a reference to a feature called people search we. Are calling it Google directory and this, is going to be available as a beta it, is going. To come both with a UI and an API so. Ass. Was, explained, by one of the panelists, you can use the query API in cloud search to retrieve results, about, people in your organization. In. Addition, to providing this API based access this, will be packaged as a an. Additional, benefit as an additional service that you get as part of G suite and we call this Google. Directory so, here is a screenshot of what you are likely to see you'll, be able to see the, both the organizational. Hierarchy that a person comes from navigate, up and down but, also we are super excited that you will be able to understand more about what. A person works on double. Click into that see who else works on the same topic areas because teams, have become far more fluid, these days and more on this will be coming on the G suite side moving forward so, we invite those of you who are interested in being part of a beta for this product at this short, link and of course take pictures but this, whole presentation will be online as well, will. Be selectively. Rolling out the beta based on the demand, we are seeing a lot of demand for this feature from customers, so we'll be rolling it out to get feedback, from a lot of you so we encourage you all to sign up for this beta with. That my, presentation, for the day is complete thank you all very much. Won, a couple, of notes for all of you to keep in mind one, is when it comes to people search. Like. Any search product, you are only going to find information that your index so. There. Is something called the Google cloud directory sync that all customers. Have access to if you're a G suite customer, you. Have to go ahead and configure that so that you index more information, because, what you can find is a function of what you have indexed in so that's one. Obvious. Point but I'm just calling it out because it's important, to note it's not like, there's no magic to finding information that's not there in the index that's yet to be invented, the, other, part here is that for those of you who are curious about the search quality metrics, we, have also published an article that works - how, Google cloud search actually does ranking, it's a standard to be a topic of general interest for a lot of people so, please go to google.com and look for Google cloud search ranking. How so on and so forth I'm sure you will find the article and. Thank. You very much for all of your interest today, we. Encourage, you all to sign, up talk, to our partners we're all here thank, you all for your time with sharing your stories thank you very much.
2019-04-14 19:23
