Sunday, November 22, 2009


Rich Leiter twittered Xyggy legal, a new beta search engine for finding similar cases on the web. Xyggy offers more than just general legal searches, so you can reach Xyggy Patent, which covers patents from 1976 onwards. Xyggy Articles searches through archived news articles of the New York Times from 1987 - 2007.

Xyggy Legal covers:
Supreme Court decisions in US. Reports volumes 1 - 544,
Federal Circuit Courts of Appeal in F.2d, volumes 178 - 999, and F.3d vol. 1 - 491
They state that they are in beta, and that they hope to complete the currently spotty coverage of recent years as they go along.

Xyggy's home page explains the Item search as follows

the fundamental idea:

Xyggy’s paradigm for search is based on the fundamental idea that search should be item-based not just text-based
In everyday life, we constantly search for and find things (items). Xyggy is bringing item search to our digital lives.

What are items?
Documents, web pages, images, music, publications, medical records, proteins, news articles, people profiles, movies, products, sounds, newswires, patents, books, customer profiles, legal cases, resumes and many more are all examples of items.

What is item search?
Finding items with a query of one or more items.

The difference
A text search returns web pages or documents containing the keywords in ranked order. Item search returns similar items in ranked order.

Text search is fine, to a point, but we live in a world of items not just keywords. Item search provides substantial advantages and additional information over keyword search. Item search is a more natural way of finding things.

Why query with one or more items?
A query with more than one item allows Xyggy to discover what the query items have in common and returns better results.
I believe they are searching not only the text but also the metadata. They have carefully selected a few databases that are made available by the government. They have also purchased a dataset from the New York Times. All three already have metadata added. Xyggy searches these in addition to the full text of the documents. This is a very interesting development. I will be interested in testing what the engine can do and will be interested in hearing back from other librarians and other users. What an interesting way to advance! We have Google Scholar's legal button and now Xyggy! Xyggy Legal went live with a demo on November 16!

The decoration is the comic strip character Ziggy, by Tom Wilson. Tip of the OOTJ hat to Rich Leiter!


Unknown said...

Hi Betsy

In the information retrieval spectrum, item-search and text-search (and related technologies such as NLP, semantic, LSA etc.) lie at opposite ends. With item-search you are literally passing an entire document as the query. In fact with Xyggy, queries can contain one or more documents and it will find all similar documents in ranked order. Xyggy is a different search paradigm and as such is not comparable to how text-search works.

Text-search with or without advanced options finds documents containing the query keywords. But, do users want to find documents with specific keywords or do they want to find similar documents? Maybe they want to do both but only the former option has been available to date.

As an aside, imagine the day when you could ask a library search system to find similar books and journal articles - Xyggy has the platform to deliver such services.

We look forward to receiving feedback from law professionals and students about Xyggy Legal.

CEO & Founder

Betsy McKenzie said...

Dinesh, you frustrate me! I want to understand what is happening beneath the hood at Xyggy. Is the algorithm working upon the entire document? Fine! But what is it doing with the damned thing? Explain it to me, I beg of you. I would love to have a conversation. I do not need or want to get to the level of code, or breaking your industrial secrets. But I am curious about how it works because when I understand that, I can make it work better for me & my patrons. I can teach it to my students better. Please explain a bit better than the opaque marketing statements you repeat here!

Unknown said...

Hi Betsy

Item-search is a marvel and I fully understand your curiosity but I have to disappoint as we cannot discuss what's going on under Xyggy's hood but I will explain how to use it.

Using Xyggy Legal, begin to enter a citation number or a portion of the named parties in the autocomplete box. A list of matching cases will appear to choose from. If the case of interest isn't in the list, enter all or more of the citation or enter more of the named parties. The list of matching cases will dynamically change as you enter more of the query.

When the case of interest appears in the list, click on it and the results will appear showing all similar cases in ranked order. Mousing-over the results will pop-up a summary window showing the citation, parties, court, docket and date information for each case. Clicking on a result will open a different browser window to display the text of the case.

At this point, the user has two options: click the 'clear' button to start a new query or continue by entering another case to the query. This is accomplished as above but underneath the first query item. In fact, you can enter multiple cases per query. The concept of finding similar cases to more than one case is something we do in our everyday lives. For example, if I'm searching a pile of case documents by hand, I'll pull out every case of interest and keep looking for similar cases. Each time an item is added to the Xyggy query the results are updated automatically. As before, clicking the 'clear' button will start a new query.

Today, a Xyggy query contains "known" items (for example, selected by citation number or parties in Xyggy Legal) but a future interface will expand the capability to help make Xyggy work even better for you.

Hope this helps.