Feeds:
Posts
Comments

Archive for June, 2008

In looking at the comments of the last post The Search for Semantic Search, I see there appears to be some interesting interpretations. Let me explain my motives, address any perceived bias and clarify my position.

Alex Iskold wrote about semantic search that we were asking the wrong questions; that it was essentially the root of the problem with semantic search engines, and; they were only capable of handling a narrow range of questions– those requiring inference. Among other things, he also wrote that his question about vocation was unsolvable; impossible, was the term he used. These ideas and the fact that Alex implied Google was a semantic search engine, and inferred that vendors must dethrone Google to be successful, motivated me to blog about it myself.

I was criticized, in the comments, for implying that the so-called “semantic search” capability of these NLP-driven search engines is weak, and due to this they do not really qualify as “semantic search” engines. Actually Kathleen Dahlgren introduced a new name in her comments: “Semantic NLP”. I was also criticized for asking a silly question and for posting my brief analysis of this one single question that Alex said was unsolvable without massively parallel computers.

Of course you cannot judge a search engine by the way it mishandles one or even a few queries. But in this case one natural language question reveals a lot about the semantic acuity of NLP, and the multiple query idea is a kind of strawman argument intended to distract us. It almost proves Alex right as it is misleading.

I do not believe people are motivated to ask wrong questions and I do not believe people ask silly questions to computer search engines while expecting a satisfactory set of results or answers. Nevertheless, when any case fails, the problem or fault does not lie with people. The search engine is supposed to inform them. The fault lies with the computer software for failing to inform. You can try and dismiss it with a lot of hand waving but just like that pesky deer fly — it is going to keep coming back.

While NLP front ends and semantic search engines are the result of millions of dollars in funding and the long work of brainiacs, and while they may be capable enough to parse sentences, build an index or other representation of terms and use some rules of grammar, they are not always accommodating or satisfying. In fact they can be quite brittle or breakable. This means they do not always work. But they do work under the right circumstances in narrowly defined situations. One of the questions here is whether they work well enough to qualify them as “semantic search” engines for English language questions.

Any vendor who comes out in public and claims they are doing “semantic search” should prove it by inferring the significance of the input with sufficient quality and acuity such that the result, or search solution or conclusion, satisfies the evidence and circumstance. This a minimum level of performance. There are tests for this. Many people use a relevance judgment as a measure of that satisfaction as far as any type of search and retrieval method or software is concerned.

With that said, my last post was about debunking the so-called complex query myth not about “testing” the capabilities of any search engine. It was about semantic search and how any search engine solves this single so-called impossible question. There were results, and they were not completely “useless” as I see, on review, that I wrote. I apologize for calling them useless.

Both Cognition and Powerset produced relevant results (with one word) that were more comprehensive than the results Google provides, in my opinion. That is not a natural language process of understanding a sentence though. Having a capacity to look up a word in a dictionary is not the same as the capacity to referentially or inferentially use the concept. In this case, to make some judgments (distinguish the significant relationships, at least) and inform the search process.

This capability to distinguish significant relationships is a key criteria of “semantic” search engines — meaning they should have a capacity to infer something significant from the input and use it. The results of this query tell a different story. You cannot just profess linguistic knowledge, call the question silly and make the reality it represents go away. This kind of problem is systemic.

As far as the so-called “semantic” search engines inferring anything siginificant from this (full sentence case) question (evidence) or circumstance of searching, I treated all the results with equal disaffirmation. What is more; I stand by that as it is supported on its face. If you look at the results of the full sentence case query at Cognition, you will notice that they are essentially the same as those from Powerset.

I reckon this could be because both engines map the parts of speech and terms from the query onto the already prepared terms and sentences from Wikipedia. This “mapping strategy” clearly fails –in this case– for some pretty obvious reasons. Without pointing out all the evidence I collected, I summed those reasons up as a lack of semantic acuity. That seems to have touched a nerve.

So I will get into the details of this below. Let me first take a moment to address the fact that one inquiry reveals all this information. Really it is not just one inquiry. It is one conceptualization. Dozens of questions can be derived by mapping from the concepts associated to these terms of this single question. For example: Where are the best careers today?; Who has the better jobs this year?; Where can I work best for now?; What occupation should I choose given the times?; etc. I tried them all and more with varying degrees of success.

One problem is that NLP practitioners are concerned with sentence structure and search engineers are concerned with indexing terms and term patterns. Either way, the methods lack a conceptual dimension and there is no apparent form of any semantic space for solving the problem. The engines have no sense of time points or coordinated space or other real contexts in which things take place. The absence of semantic acuity is not something that only affects a single inquiry. It will infect many inquiries just as a disease infects its host.

Now that I recognize the problem, if I were challenged to a wager, I would wager that I could easily produce 101 regular English language questions that would demonstrate this affliction. The search engines may produce a solution, except that the results would be mostly nonsense and not satisfying. It would prove nothing more and nothing less than I have already stated. What say you Semantic Cop?

I should mention that I have long suspected that there was a problem mapping NLP onto a search process and I could not put my finger on it. A literature search on evaluations of text retrieval methods will show, in fact, that the value of part of speech processing (in text search and retrieval) has long been regarded as unproven. By taking the time to investigate Alex Iskold’s complex query theory I gained more insight into the nature and extent of this problem. It is not just a problem of finding a definition or synonyms for a given term as some reader’s may infer. Let me explain further.

While Powerset, Cognition and Hakia each had the information that a vocation was a kind of altruistic occupation, and the search circumstance (a hint) that the information seeker could be looking for an occupational specialty or career, they did not really utilize that information. The failure, though, really wasn’t with their understanding of the terms occupation or vocation. Their failure was specifically related to the NLP approach to the search process. That is supported by the fact that these different search products employing NLP fail in the same way.

That should not be taken to mean that the products are bad or useless. Quite to the contrary, the product implementations are really first class productions and they appear to improve the user experience as they introduce new interface techniques. I think NLP technology will continue to improve and will eventually be very useful, particularly at the interface as Alex noted in his post. But does that make them semantic search engines?

Lest I have been ambiguous, let me sum up and clarify by referring back to the original question: Whether you are looking at Powerset, Cognition or Hakia results, they clearly did not understand the subordinate functionality of the terms /best/, /vocation/, /me/ and /now/ in the sentence.

They clearly could not conceptualize ‘best vocation’ or ‘now’– they could only search for those keyword patterns in the index or data structures created from the original sentences. That is not just ‘weak’ semantics that is not semantic search at all. Maybe they “understood” the parts of speech but they did not infer the topic of inquiry nor did they properly map the terms into the search space. Google did not fare any better in this case, but Google does not claim to be a semantic search engine. So where’s the semantics?

By that I mean (for example) that interpreting /now/ from the natural language question ‘what is the best vocation for me now’ as an adverb, does not improve the search result. Treating it as a keyword or arbitrary pattern does not improve the search result. And it demonstrates a clear and present lack of acuity and understanding of the circumstance.

Finding the wrong sense of /now/ and showing it is of dubious value. An inference from /now/ to ( –> ) today, at present, this moment in time, or to this year or age, and using that as evidence leading to an informed conclusion would demonstrate some semantic acuity in this case. Most people have this acumen, these NLP search engines obviously do not–according to the evidence in this case.

The NLP vendors defend this defect by accusing people of not asking the right question in the right way or not asking enough questions. That is like me saying to my wife:

If you want some satisfying information from me you better use a lot of words in your question and it better not be silly. Don’t be too precise and confuse me and don’t use an idiom and expect me to satisfy you. I’ll still claim to understand you. It is you that asks silly questions. That not being enough, you also have to nag me with more long and hard questions before you say my responses are rubbish.

If I should either desire or dare to do that at all, what do you think her response would be? More importantly what do prospects say when you tell them their questions are silly?

I do not need to proceed with a hundred questions when with a dozen or so I have enough evidence to deduce that these NLP-driven search engines are limited when it comes to ** inferring the topic of inquiry **. In some cases they are simply unable to draw on the *significant structures or patterns of input, evidence or circumstance” and produce a suitable solution.

What bothers me is that some of these so-called “semantic search engines” claim to “understand” in general. I did that too, a very long time ago. Yes, I was there in the back of the room at DARPA and NIST meetings and I have been at PARC and the CSLI for presentations. I was challenged then. And it enlightened me. If such claims go unchallenged it will only serve to demean the cultural arts and independent critical thinking and confuse prospects about the capabilities regular people expect of semantic products. I do not wish to lower the bar.

In this instance, and there are many similar cases that could be derived from the semiotic characteristics of this instance, the NLP-driven engines do not show the slightest acumen for inferring the topic of inquiry. I hope the discerning reader sees that it is not just about some synonyms. If they could infer the topic of inquiry, that would demonstrate a little understanding… at least a capacity to learn.

The result, in all such cases, is that these so called “reasoning-engines” and semantic search engines do not lead us to a satisfying consequence or conclusion at all. They have technical explanations such as synonymy and hyponymy for any word, yet, if the software cannot infer the sense of everyday terms, is it even sensible to call the methods “semantic”? Just because the vendors profess linguistic knowledge does not mean their their semantics are any more than just another marketing neologism.

It may be called semantic NLP but that does not qualify as semantic search in my opinion.

Read Full Post »

In a recent Read Write Web article that was much more myth than reality, Alex Iskold posits the fact that a semantic search engine must dethrone Google (myth1). Fortunately by the end of his article he concludes that he was mislead into thinking that. I do not think he was misled at all. I just think he is confused about it all.

He posits a few trivial queries (myth 2) to show that there is “not much difference” between Powerset and Hakia and Freebase (myth 3). And that semantic search “is no better than Google search ” (myth 4). After that Alex writes that there is a set of problems suitable to semantic search. He says these problems are wonderfully solved (already) by relational databases (myth 5).

It makes one wonder why we should mess with semantic search if the problems are already solved. It is not true. That is why. Neither was any of the talk about query complexity, true.

It is not all these myths, exactly, but unclear thinking that leads to false expectations as well as false conclusions. Alex seems to be confused about the semantic web and semantic search. These are two different things but somehow Alex morphs them into one big database. Because I do want this post to impart some information instead just being critical of a poorly informed post, let me start by debunking the myths.

Myth1: Semantic Search should dethrone Google

For many search problems, semantics plays no role at all. Semantics plays a very limited role when the query is of a transactional nature, e.g., a search problem of the type: find all x.

Google is a search engine that solves search problems of this type. Yet the Google kingdom is based on being a publisher. Google uses super-fast and superficial keyword search to aggregate dynamic content from the internet for information seekers and advertisers alike. Google’s business does not even lend itself to semantic search for some very obvious reasons having to do with speed and scalability. Google’s best customers know exactly what they want and they certainly do not want any “intellectual” interpretations.

None of the so-called semantic search engine companies, that I know of, are pursuing a business strategy to dethrone Google as an information-seeker’s destination of choice. Powerset, for example, is not aggregating dynamic content like Google. It’s business model does not seem to be based on a publishing or advertising model.

Powerset is using their understanding of semantics to assist the user (of wikipedia) in relating to that relatively static content, from several different mental or rational and conceptual perspectives. This is meant to assist the information-seeker with interpreting the content. That is a good and valid application of semantics.

This is not the position a company seeking to unseat Google would take. A company seeking to unseat Google would be better positioned by producing technology to assist advertisers in classifying, segmenting and targeting buyers.

Myth 2: Trivial and Complex Queries

Unfortunately Alex did not supply any complex examples in his post. He tried to imply that his trivial queries were complex and the most complex was impossible to solve. This query was the one labeled impossible: “What’s the best vocation for me now?” I will use Alex’s query to debunk his misguiding assumptions. First, let’s clarify by looking at the search problems represented by the Alex’s natural language queries.

Note 1: Alex offers the first query as impossible to solve. It must be because Alex is expecting a machine and some software to divine his calling based presumably on his mood now and some mind-reading algorithm. I should hope most people would seek a human counselor rather than rely on the consul of a semantic search engine for addressing their calling. It is fair to use a search engine to find a career or occupation and it is valid to expect a semantic search engine to “understand” the equivalence relationship between the terms occupation and vocation, in this context.

As I suggested best + vocation, or just vocation alone is a simple solution that should be easy to satisfy. However, this simple search solution fails on all search engines. Even so-called semantic search engines have a problem with this query (see comparative search results under myth 4 below). It is not because it is complex query. It is because Alex used the word vocation. This word is not frequent and search engines do not know its synonyms. This is a complex concept as it takes semantic acuity to “understand” it. No one talks about semantics in terms of acuity though.

Nonetheless, a search for vocation + best, and sorting the results by most recent date, will however, create a valid search context in which one can reasonably expect a solution from their semantic search engine. Most people, I am assuming, would have a more reasonable expectation than Alex; one that may be fulfilled by this internet page suggested by Readware:

A semantic search engine needs semantic acuity to “understand” that the concept of a vocation and the concept of an occupation are related. Obviously none of the search engines mentioned in Alex’s article have such acuity of understanding. Some of the search engines tried to process the pronoun me and the word now. Instead of being a solution, it created a problem as can be seen in the search results (under myth 4) below.

Note 2: This query needs a search engine with some more exotic search operators than a simple keyword search engine might provide. The query, however, is not complex. Some search engine may index US Senator as a single item to facilitate such a search. A search engine would need extended Boolean logic to process phrases using a logical AND between them. A more seasoned search engine, such as Google, would parse and logically process the query from a search box, without any specifying logic, and return an acceptable result. NLP-based engines (like Hakia and Powerset) try to do this too. They use propositional logic instead of Boolean logic. The effects are not very satisfying as can be seen below (in the search results listed under myth 4).

A more sophisticated and indeed “semantic” search engine may interpret foreign entity according to a list of “foreign entities”. It would take some sophisticated semantics to algorithmically interpret what type of labels may be subordinate to foreign entity. For example: A German businessman, a Russian bureaucrat, a Japanese citizen, an American taxpayer. Which is the foreign entity?

Yet, it is also clear that an inventory of labels can be assembled under certain contexts. Building such an inventory constitutes the building of knowledge. A semantic search engine should help inventory and utilize knowledge for future researches. None of the semantic search engines that Alex mentioned do anything like this. Readware does do this.

Note 3: This search would benefit from a search engine that recognizes names. I think Hakia has an onomastic component. I am not sure about Powerset. However, this search works on nearly any search engine because their are plenty of pages on the web that contain the necessary names and terms. Otherwise there is nothing complex about this query.

The reality, as you can see, is that every query Alex offered is trivial. Yet it demonstrates what is wrong with so-called “semantic search”. That is, today’s semantic search products, including the NLP -powered search engines masquerading as “semantic” search, fail at real tests of semantic acuity. Before I get into the evidence though. Let me just say something about semantic search technologies in general.

Semantic Search Technologies

There are no public semantic search engines today. There are search engines and there are search engines with Natural Language Processors (NLP) that work on the indexing and query side of the search equation. Whether or not databases are used to store processed indexes or search results, databases and database technology like RDBMS and SQL have nothing to do with it.

The search engines that have the capacity for natural language processing usually claim to “understand” word and/or sentence semantics– in a linguistic sense. This usually means that they understand words by their parts of speech, or they can look up definitions in a resource. Hakia and Powerset fall into this class, as does Cognition and several other search engines both in the U.S. and abroad. These are called semantic search engines and they claim to understand word sense and do disambiguation and so forth and so on, but as I will show below: at questionable acuity.

Google is not a semantic search engine at all. While Hakia and Powerset may represent some small part of the spectrum of semantic search engines they are hardly representative of semantic search. Along with Freebase and Powerset, more representative of “semantic web” search is SWSE, Swoogle and Shoe.

Besides these semantic web search engines, there are semantic search engines akin to Hakia, such as Cognition, as mentioned in this article at GigaOM, along my own favorite Readware. So, in summary, Alex’s comparison is not representative and is really poor evidence.

Myth 3: No difference between Powerset and Hakia and Freebase.

Well this is just ridiculous. It is not only a myth, it is pure misinformation. Nothing could be further from the truth. While Powerset and Hakia use NLP technology that could be construed as similar, Freebase is a essentially an open database that can be queried in flexible ways. Freebase and Powerset happen to be somewhat comparable because Powerset works on Wikipedia and uses RDF to store data, and semantic triples (similar to POS) to perform some reasoning over the data. Freebase also stores Wikipedia-like data in RDF tuples.

It is probably also worthwhile to mention that Hakia’s NLP comes from the long time work and mind of the eminent professor Victor Raskin and his students. Powerset’s NLP comes from the work of Ronald Kaplan, Martin Kay and others connected with Palo Alto Research Center, Stanford University and the Center for the Study of Language and Information (CSLI). Cognition’s technology is based on NLP work done by Kathleen Dahlgen.

While Hakia, Powerset and Cognition represent these notable NLP approaches, their search methods and results show they do not know a great deal about search tactics and solutions. They do not seem to be successful in mapping the sentence semantics into more relevant and satisfying search results. It seems, from the evidence of these queries, they only know how to parse a sentence for its subject, object and verb and, a lot like Google, find keywords.

Myth 4: Semantic Search is No Better than Google.

Hakia and Powerset are like neophytes in Google’s world of search. That alone makes these engines no better than Google. Yet, that does not apply to semantic search in general. The truth is that the semantics of the search engines we are talking about (Hakia, Powerset, Freebase and Search Monkey), do not appear to make the results any worse than those from Google. Let’s take a look at the Google search results for ‘What is the best vocation for me now’:

As may be predicted, the results are not very good (because the keyword vocation is not popular). Google also wants to be sure we do not mean ‘vacation’ instead of vocation. Hakia , on the other hand , strictly interpreted the query:

Just like the results from Google, these are not very satisfying. You might think that because Hakia is a semantic search engine, it would have the semantic acuity to “understand” that vocation and occupation are related. As you can see in the following search result, this could not be farther from the truth:

Not one of Hakia’s results had to do with occupational specialties or opportunities for career training and employment. Powerset did not produce any results when the term vocation is used and it really had nothing on occupation so it searched for best + me + now. There is nothing semantic about that and it is a pretty bad search decision as well. The results are useless; I will post them so you can judge for yourself:

When you have results like this, it really does not matter what kind of user interface you have to use. If it is a bad or poor user interface, it only makes the experience that much worse. Even if it is a good, fancy, slick or useful interface, it won’t improve irrelevant results.

Another so-called semantic search engine, Cognition, did not fare any better:

This above search result is useless provides a starting point for further investigation, as is does the search for occupation:

I actually was mildly surprised that Cognition related the term occupation to the acronym MOS, which means Military Occupational Specialty. Then I saw that they did not distinguish the acronym from other uses of the three letter keyword combination. Again not a very satisfying experience. I did not leave Freebase out, I just left them until last. All Freebase results do is confirm that vocation is an occupation or a career:

It was not possible for freebase to accept and process the full query. As this result shows, the data indicates that a vocation is also known as an occupation but none of these engines realize that fact.

Myth 5: Already solved by RDBMS.

If the search problem or the “semantic” problem could be solved by the RDBMS, Oracle would be ten times the size of Google and Google might be using Oracle’s technology if it existed at all. None of these problems (aggregated search, semantic parsing of the query and text, attention, relevance) are solved by any RDBMS. But Alex brushed over the real problems to make the claim that it is all up to the User Interface and the semantics only matter there. I suppose that was the point he was trying to make by including Search Monkey in his comments. This is just hog wash though. By that I mean that it is not true and it is in fact misleading.

Conclusions

It is plain to see that a semantic search engine needs acuity to discern the differences and potential relations that can form between existing terms, names and acronyms. It is also plain to see that none of the commercial crop of search engines have it. The Natural Language search engines, which have dictionaries as resources, do not associate vocation to occupation (for example) and therefore cannot offer any satisfying search results.

There are 350,000 words in the English language. How many do you suppose are synonymous and present a case just like this example? Parsing a sentence for its subject, object and verb, is fine. It does not mean it will be useful or helpful in producing satisfying search results.

It is foolish to think that NLP will be all that is needed to obtain more relevant search results. The fact is that search logic and search tactics are arts that are are largely unrelated to linguistics or NLP or database technology. While language has semantics, testing of the semantics of so-called semantic search engines has demonstrated that the semantics, if they are semantics, are pretty weak. I have demonstrated that semantic acuity plays a large role in producing more relevant and satisfying search results. A semantic search engine should also help inventory and utilize knowledge for future researches. An informed search produces more satisfying results.

Read Full Post »