Semi-supervised learning of geographical gazetteer from the internet

Semi-supervised learning of geographical gazetteers from the internet Olga Uryupina Computational Linguistics, Saarland University Building 17 Postfach 15 11 50 66041 Saarbr¨ucken, Germany ourioupi@coli.uni-sb.de Abstract In this paper we present an approach to the ac- quisition of geographical gazetteers. Instead of creating these resources manually, we propose to extract gazetteers from the World Wide Web, using Data Mining techniques. The bootstrapping approach, investigated in our study, allows us to create new gazetteers using only a small seed dataset (1260 words). In addition to gazetteers, the system pro- duces classifiers. They can be used online to determine a class (CITY, ISLAND, RIVER, MOUNTAIN, REGION, COUNTRY) of any geographical name. Our classifiers perform with the average accuracy of 86.5%. 1 Introduction Reasoning about locations is essential for many NLP tasks, such as, for example, Information Extraction. Knowledge on place names comes normally from a Named Entity Recognition module. Unfortunately, most state-of-the-art Named Entity Recognition systems sup- port very coarse-grained classifications and thus can dis- tinguish only between locations and non-locations. One of the main components of a Named Entity Recog- nition system is a gazetteer — a huge list of preclas- sified entities. It has been shown in (Mikheev et al., 1999) that a NE Recognition system performs reason- ably well for most classes even without gazetteers. Lo- cations, however, could not be reliably identified (51,7% F-measure without gazetteers compared to 94,5% with a full gazetteer). And obviously, when one needs more so- phisticated classes, including various types of locations, gazetteers should become even more important. One possible solution would be to create gazetteers manually, using World atlases, lists of place names on the Web, and already existing digital collections, such as (ADL, 2000). This task is only feasible, of course, when those resources have compatible formats and, thus, can be merged automatically. Otherwise it becomes very time- consuming. Manually compiled gazetteers can provide high- quality data. Unfortunately, these resources have some drawbacks. First, some items can simply be missing. For example, the atlases (Knaur, 1994), (Philip, 2000), and (Collins, 2001), we used in our study, do not list small islands, rivers, and mountains. Such gazetteers contain only positive information: if � is not classified as an ISLAND, we cannot say whether there is really no is- land with the name �, or simply the gazetteer is not complete. Another problem arises, when one wants to change the underlying classification, for example, subdi- viding CITY into CAPITAL and NON-CAPITAL. In this case it might be necessary to reclassify all (or substantial part of) the items. When done manually, it again becomes a very time-consuming task. Finally, geographical names vary across languages. It takes a lot of time to adjust a French gazetteer to German, moreover, such a resource can hardly bring a lot for languages with non-Latin al- phabets, for example, Armenian or Japanese. Even col- lecting different variants of proper names in one language is a non-trivial task. One possible solution was proposed in (Smith, 2002). At least some information on …

Content truncated to 500 words