Mythic Search: Yext Introduces the Phoenix with Summer Updates
September 15, 2021
Enterprise search firm Yext is launching new features and a revamped algorithm, poetically named “Phoenix.” We learn about the updates from the press release, “New Yext Features and Algorithm Update Bring AI Search Optimizations to Businesses” at PR Newswire. We learn:
“In addition to features powered by Phoenix like dynamic reranking, the release introduces revamped test search and experience training, as well as a reimagining of Yext’s data connector and app frameworks — all to equip businesses with modern and powerful search solutions.”
The dynamic reranking feature sounds promising. Phoenix analyzes user behavior to push the most relevant results to the top. We are given an example:
“If customers consistently click on a blog post when searching for vaccine information on a healthcare organization’s website, dynamic reranking will push that content to the top of the search results page so it appears first any time someone searches about vaccines. The Phoenix update also introduces more relevant results for queries about locations that are ‘open now’ and rich text fields, like lists, in featured snippets.”
Another feature is the ability to build Yext platform configurations and package them into installable apps. The update also makes it easy to test search experiences from the customer’s point of view. But Yext may promise a bit much with its updates to data connectors:
“With the new update to Yext’s data connectors framework, businesses can use a low-code ‘extract, transform, load’ (ETL) tool that extracts all of their data and transforms it into the same format for easy integration into their knowledge graph (a unique brain-like database of facts).”
We do not want to be critical, but we are skeptical when a vendor of search and retrieval uses the word “all.” Certain types of data are notoriously difficult to access, like chemical structures, audio, video, images, and product-management quality assurance data, to name a few. Retrieving “all” data is unlikely at prices most organizations can afford. Still, it does sound like Phoenix is a step forward from the company that promises “Search made for today. Not 1999.” Today’s “search” dates back a half century, but who is interested in history?
Cynthia Murrell, September 15, 2021
Coveo: A Search Vendor Repositions, Pivots, and Spins
September 13, 2021
Coveo was a vendor of search and retrieval software. Then Coveo morphed into help desk and self-service software. Now the company appears to be spinning like a whirling dervish into a new positioning. “Coveo Adds More Developer Features to Its AI Powered Digital Experience Platform” explains:
Coveo Solutions Inc., a unicorn startup that helps companies such as Salesforce.com Inc. and Adobe Inc. improve their websites with artificial intelligence, today introduced new features to help developers more easily use its technology.
A couple of minor points. Coveo has ingested about $330 million since it was set up in 2005. I think that works out to 16 years, which in my experience makes Coveo something other than a start up. Your book may be different, of course.
I am not into enterprise search, but I find it interesting that this company is spinning in an AI powered digital experience platform. I don’t have a clue how to define “artificial intelligence.” I simply don’t know what a “digital experience platform” is.
That may not matter. The point is keep moving, changing, and morphing in order to generate sufficient revenue to make long suffering investors happy campers and differentiate the commodity of search technology from open source and proprietary options.
Oh, do dervishes get dizzy? I do.
Stephen E Arnold, September 13, 2021
Interesting Number: Apple Sells Access
September 3, 2021
I read “Google to Pay Apple $15 Billion to Remain Default Safari Search Engine in 2021.” The write up states:
It’s long been known that Google pays Apple a hefty sum every year to ensure that it remains the default search engine on iPhone, iPad, and Mac. Now, a new report from analysts at Bernstein suggests that the payment from Google to Apple may reach $15 billion in 2021, up from $10 billion in 2020. In the investor note, seen by Ped30, Bernstein analysts are estimating that Google’s payment to Apple will increase to $15 billion in 2021, and to between $18 billion and $20 billion in 2022.
Apple and Google care about their users and their “experience.” That’s a mellifluous thing to say, particularly in an anti-trust deposition.
Let’s put the allegedly accurate number in context:
The metasearch engine DuckDuckGo may be in the $70 million range. That is in the neighborhood of 200 times the metasearch system’s estimated revenues for 2020.
Stephen E Arnold, September 3, 2021
Wiki People: One Cannot Find Online Information If It Is Censored
September 2, 2021
Women have born the brunt of erasure from history, but thanks web sites like Wikipedia, their stories are shared more than ever. There is a problem with Wikipedia though, says CBC in the article: “Canadian Nobel Scientist’s Deletion From Wikipedia Points To Wider Bias, Study Finds.” Wikipedia is the most comprehensive, collaborative, and largest encyclopedia in human history. It is maintained by thousands of volunteer editors, who curate the content, verify information, and delete entries.
There are different types of Wikipedia editors. One type is an “inclusionist,” an editor who takes broad views about what to include in Wikipedia. The second type are “deflationists,” who have high content standards. American sociologist Francesca Tripodi researched the pages editors deleted and discovered that women’s pages are deleted more than men’s. Tripodi learned that 25% of women’s pages account for all deletion recommendations and their pages only make up 19% of the profiles.
Experts say it is either gender bias or notability problem. The notability is a gauge Wiki editors use to determine if a topic deserves a page and they weigh the notability against reliable sources. What makes a topic notable, Tripodi explained, leads to gender bias, because there is less information on them. It also does not help that many editors are men and there are attempts to add more women:
“Over the years, women have tried to fix the gender imbalance on Wikipedia, running edit-a-thons to change that ratio. Tripodi said these efforts to add notable women to the website have moved the needle — but have also run into roadblocks. ‘They’re welcoming new people who’ve never edited Wikipedia, and they’re editing at these events,’ she said. ‘But then after all of that’s done, after these pages are finally added, they have to double back and do even more work to make sure that the article doesn’t get deleted after being added.”
Unfortunately women editors complain they need to do more work to make sure their profiles are verifiable and are published. The Wikipedia Foundation acknowledges that the lack of women pages, because it reflects world gender biases. The Wikipedia Foundation, however, is committed to increasing the amount of women pages and editors. The amount of women editors has increased over 30% in the past year.
That is the problem when there is a lack of verifiable data about women or anyone erased from history due to biases. If there is not any information on them, they cannot be searched even by trained research librarians like me. Slick method, right?
Whitney Grace, September 2, 2021
Amazon Search: Just Outstanding
September 2, 2021
Authors at Paste Magazine are dedicated to assembling lists of the best streaming content from Netflix, Hulu, Amazon Prime, and other services. They know almost as much about these content libraries as their developers. The title in Paste Magazine’s article, “Amazon Prime Video’s Library Is Not Genuinely Impossible To Browse” says it all.
It is notoriously difficult to browse Amazon Prime’s content library and the problem was noted in 2018. Amazon Prime’s library contains a lot of content, much of it is considered unwatchable. The only way to locate anything is searching by its proper name, but users who want to browse films like physicals libraries and video stores of yore are abandoned.
Amazon Prime has also hidden its search function, instead it wants users to work around this road block:
It quickly becomes apparent that there is no obvious way to view that full list of sci-fi movies, suggesting that Amazon doesn’t want consumers to be able to easily find that kind of information—its user experience is built around you choosing one of the small handful of suggested films, or knowing in advance what you want to see and then specifically searching it out. However, it is possible to see the full list—in order for it to display, you just have to click on any specific sci-fi film, look at the movie’s genre tags, and click on the words “science fiction” once again.”
The search function is worse than that available in a medieval scriptorium. When users return to certain genre pages and browse the supposed complete list, the same twenty-one movies continuously reload.
Amazon Prime has thousands of titles and is designed by a high tech company, yet it cannot fix its search function? Why does Amazon, an important company that is shaking the film and television industry, not offering its users the best of the best when it comes to search? Amazon did A9, it sucked in Lucid Imagination “experts,” it intruded on Elastic search territory. And now search doesn’t work the way users expect. Has another high-tech outfit become customer hostile or just given up making search useful?
Whitney Grace,September 1, 2021
Semantic: Scholar and Search
September 1, 2021
The new three musketeers could be named Semantic, Scholar, and Search. What’s missing is a digital d’Artagnan. What are three valiant mousquetaires up to? Fixing search for scholarly information.
To learn why smart software goes off the rails, navigate to “Building a Better Search Engine for Semantic Scholar.” The essay documents how a group of guardsmen fixed up search which is sort of intelligent and sort of sensitive to language ambiguities like “cell”: A biological cell or “cell” in wireless call admission control. Yep, English and other languages require context to figure out what someone might be trying to say. Less tricky for bounded domains, but quite interesting for essay writing or tweets.
Please, read the article because it makes clear some of the manual interventions required to make search deliver objective, on point results. The essay is important because it talks about issues most search and retrieval “experts” prefer to keep under their kepis. Imagine what one can do with the knobs and dials in this system to generate non-objective and off point results. That would be exciting in certain scholarly fields I think.
Here are some quotes which suggest that Fancy Dan algorithmic shortcuts like those enabled by Snorkel-type solutions; for example:
Quote A
The best-trained model still makes some bizarre mistakes, and posthoc correction is needed to fix them.
Meaning: Expensive human and maybe machine processes are needed to get the model outputs back into the realm of mostly accurate.
Quote B
Here’s another:
Machine learning wisdom 101 says that “the more data the better,” but this is an oversimplification. The data has to be relevant, and it’s helpful to remove irrelevant data. We ended up needing to remove about one-third of our data that didn’t satisfy a heuristic “does it make sense” filter.
Meaning: Rough sets may be cheaper to produce but may be more expensive in the long run. Why? The outputs are just wonky, at odds with what an expert in a field knows, or just plain wrong. Does this make you curious about black box smart software? If not, it should.
Quote C
And what about this statement:
The model learned that recent papers are better than older papers, even though there was no monotonicity constraint on this feature (the only feature without such a constraint). Academic search users like recent papers, as one might expect!
Meaning: The three musketeers like their information new, fresh, and crunchy. From my point of view, this is a great reason to delete the backfiles. Even thought “old” papers may contain high value information, the new breed wants recent papers. Give ‘em what they want and save money on storage and other computational processes.
Net Net
My hunch is that many people think that search is solved. What’s the big deal? Everything is available on the Web. Free Web search is great. But commercial search systems like LexisNexis and Compendex with for fee content are chugging along.
A free and open source approach is a good concept. The trajectory of innovation points to a need for continued research and innovation. The three musketeers might find themselves replaced with a more efficient and unmanageable force like smart software trained by the Légion étrangère drunk on digital pastis.
Stephen E Arnold, September 1, 2021
Enterprise Search: What Is Missing from This List
August 31, 2021
I got a wild and wooly announcement from something called The Market Gossip. The message was that a new report about enterprise search has been published. I never heard of the outfit (Orbis Research) in Dallas.
Take a look at this list of vendors covered in this global predictive report:
Notice anything interesting? I do. First, Elastic (commercial and open source) is not in the list. Second, the Algolia system (a distant cousin of Dassault Exalead) is not mentioned. Weird, because the company got another infusion of cash.) Three, the name of LucidWorks (an open source search recycler) is misspelled. Fourth, the inclusion of MarkLogic is odd because the company offers an XML data management system. Sure, one can create a search solution but that’s like building a real Darth Vader out of Lego blocks. Interesting but of limited utility. Fifth, the inclusion of SAP. Does the German outfit still pitch the long-in-the-tooth TREX system? Sixth, Microsoft offers many search systems. Which, I wonder, is the one explored?
Net net: Quite a thorough research report. Too bad it is tangential to where search and retrieval in the enterprise is going. If the report were generated by artificial intelligence, the algorithm should be tweaked. If humans cooked up this confection, I am not sure what to suggest. Maybe starting over?
Stephen E Arnold, August 30, 2021
Algolia: Now the Need for Sustainable, Robust Revenue Comes
August 27, 2021
We long ago decided Algolia was an outfit worth keeping an eye on. We were right. Now Pulse 2.0 reports, “Algolia: $150 Million Funding and $2.25 Billion Valuation.” The company closed recently on the Series D funding, bringing its total funding to $315 million. Putting that sum to shame is the hefty valuation touted in the headline. Can the firm live up to expectations? Reporter Annie Baker writes:
“This latest funding round reflects Algolia’s hyper growth fueled by demand for ‘building block’ API software that increases developer productivity, the growth in e-commerce, and digital transformation. And this additional investment enables Algolia to scale and serve the increased demand for the company’s Search and Recommendations products as well as fuel the company’s continued product expansion into adjacent markets and use-cases. … This new funding round caps a landmark year that saw significant growth and product innovation. And Algolia launched with the goal of creating fast, instant, and relevant search and discovery experiences that surfaced the desired information quickly. Earlier this year, the company had announced its new vision for dynamic experiences, advancing beyond search to empower businesses to quickly predict a visitor’s intent on their digital property in real time, in the session, and in the moment. And the business, armed with this visitor intent, can surface dynamic content in the form of search results, recommendations, offers, in-app notifications, and more — all while respecting privacy laws and regulations.”
Baker notes Algolia’s approach is a departure from opaque SaaS solutions and monolithic platforms. Instead, the company works with developers to build dynamic, personalized applications using its API platform. Over the last year and a half, Algolia also added seven new executives to its roster. Headquartered in San Francisco, the company was founded in 2012.
Cynthia Murrell, August 27, 2021
Google Fiddled Its Magic Algorithm. What?
August 19, 2021
This story is a hoot. Google, as I recall, has a finely tuned algorithm. It is tweaked, tuned, and tailored to deliver on point results. The users benefit from this intense interest the company has in relevance, precision, recall, and high-value results. Now a former Google engineer or Xoogler in my lingo has shattered my beliefs. Night falls.
Navigate to “Top Google Engineer Abandons Company, Reveals Big Tech Rewrote Algos To Target Trump.” (I love the word “algos”. So colloquial. So in.) I spotted this statement:
Google rewrote its algorithms for news searches in order to target #Trump, according to target Trump, according to @Perpetualmaniac #Google whistleblower, and author of the new book, “Google Leaks: An Expose of Bit Tech Censorship.”
The write up states:
As a senior engineer at Google for many years, Zach was aware of their bias, but watched in horror as the 2016 election of Donald Trump seemed to drive them into dangerous territory. The American ideal of an honest, hard-fought battle of ideas — when the contest is over, shaking hands and working together to solve problems — was replaced by a different, darker ethic alien to this country’s history,” the description adds. Vorhies said he left Google in 2019 with 950 pages of internal documents and gave them to the Justice Department.
Wowza. Is this an admission of unauthorized removal of a commercial enterprise’s internal information?
The sources for this interesting allegation of algorithm fiddling are interesting and possibly a little swizzly.
I am shocked.
The Google fiddling with precision, recall, objectivity, and who knows what else? Why? My goodness. What has happened to cause a former employee to offer such shocking assertions.
The algos are falling on my head and nothing seems to fit. Crying’s not for me. Nothing’s worrying me. Because Google.
Stephen E Arnold, August 19, 2021
Milvus and Mishards: Search Marches and Marches
August 13, 2021
I read “How We Used Semantic Search to Make Our Search 10x Smarter.” I am fully supportive of better search. Smarter? Maybe.
The write up comes from Zilliz which describes itself this way: The developer of Milvus “the world’s most advanced vector database, to accelerate the development of next generation data fabric.”
The system has a search component which is Elasticsearch. The secret sauce which makes the 10x claim is a group of value adding features; for instance, similarity and clustering.
The idea is that a user enters a word or phrase and the system gets related information without entering a string of synonyms or a particularly precise term. I was immediately reminded of Endeca without the MBAs doing manual fiddling and the computational burden the Endeca system and method imposed on constrained data sets. (Anyone remember the demo about wine?)
This particular write up includes some diagrams which reveal how the system operates. The diagrams like the one shown below are clear, but I
the world’s most advanced vector database, to accelerate the development of next generation data fabric.
The idea is “similarity search.” If you want to know more, navigate to https://zilliz.com. Ten times smarter. Maybe.
Stephen E Arnold, August 13, 2021
-
- Subscribe to Beyond Search
Feature archive
News archive
-
