metadata | Advocate's Studio

I cannot wait for this tool to hit the public! As Studio readers know, I love relevance, personalization and curation of the vast information available on the Web. This is because I have a burning desire to be able to put my finger immediately on whatever I might be looking for and get information delivered to me that I want, even before I know that I want it. Right now, my main tools for that are my6sense (Google Reader and social streams) and Feedly (Google Reader).

While I feel like I have a serviceable handle on the information flow, Twitter still eludes me somewhat. The my6sense Chrome extension helps dramatically when I am on the Web on my computer. But apart from filtering by age of tweet, there is not a great deal of granularity or customization, apart from what happens behind the scenes when I interact with the information it serves up.

Coming soon to a Twitter near you is a very cool app called Bottlenose. It will be released in closed beta in about a month and then introduced to the public at some time thereafter. Bottlenose is aimed directly at Twitter with the goal of highlighting for you the information you deem most important. The app works both behind the scenes and with your voluntarily categorized descriptive data (you can tag your tweets with topics when you make them). The app assigns a score to tweets based on a variety of factors (i.e. Klout scores, relevance, etc.) and does its own tagging of tweets with metadata categories. On this basis, you can filter the information you are receiving by narrowing in on the category you are looking for and determine the weight you should give a particular tweet based on its score. Additionally, the app reads your own tweet history to get a sense of what you are interested in and will offer back relevant information in line with those interests. All in real-time.

Current iteration of the app is web only, but the fine folks at Bottleneck are looking to expand to mobile applications, generally upping the anticipated quality of my checkout line reading accordingly.

I cannot wait. I already have my invite request in – you can too here.

Hat tip to MG Siegler over at Techcrunch.

“Search” is familiar territory for lawyers – the term impliedly describes computer-aided discovery, evolved from the longer, two-syllable term “research”, which, of course, includes old school book-based methods. But search itself is an evolving concept. We are far from the days of early Boolean search on a Walt II terminal. Furthermore, electronic search for lawyers now must adress the needs of both the legal researcher and the e-discovery expert.

I stumbled across an article today at Law.com’s Law Technology News entitled “What Lawyers Need To Know About Search Tools.” The article is written byMark Grossman and Terry Sweeney. Glancing only at the title, I thought it might be a *yawn* list of free on-line legal resources or *yawn* 1L explanation of how to form a search query. Not so. Written from the e-discovery perspective with the apparent intent of helping lawyers properly respond to discovery requests and comply with the Rules of Civil Procedure, the article is a decent critique of keyword searching as an effective means of information retrieval, particularly over vast stores of e-documents.

By far, the best part of the article is the section entitled A VARIETY OF MODELS, which discusses variations on the theme of concept searching. The authors explain how taxonomies and ontologies (oooh, semantic!) can power-up search systems, retrieving far more relevant results than simple keyword search, particularly if beefed up with mathematical or statistical algorithms and/or combined with metadata. The authors concede the keyword and Boolean searching is still useful for small and rolling data sets. Nonetheless, it is clear that new search models are pushing our ability to quickly find and categorize information and changing the very definition of what constitutes a “reasonable” search.

The authors conclude that perfection is not the goal and reasonableness is, well, reasonable in e-discovery. My take away is a little different – the better our machine language / learning systems, tighter the algorithms and more consistent and deep the metadata, the closer we will get to perfection in search results. I, for one, can’t wait.