22.3.3 Event Detection and Analysis
The task of event detection and classification is to identify mentions of events in texts and then assign those events to a variety of classes. For the purposes of this task, an event mention is any expression denoting an event or state that can be assigned to a particular point, or interval, in time. The following markup of Example 22.3.1 shows all the events in this text.
[EVENT Citing] high fuel prices, United Airlines [EVENT said] Friday it has [EVENT increased] fares by $6 per round trip on flights to some cities also served by lower-cost carriers. American Airlines, a unit of AMR Corp., immediately [EVENT matched] [EVENT the move], spokesman Tim Wagner [EVENT said]. United, a unit of UAL Corp., [EVENT said] [EVENT the increase] took effect Thursday and [EVENT applies] to most routes where it [EVENT competes] against discount carriers, such as Chicago to Dallas and Denver to San Francisco.
In English, most event mentions correspond to verbs, and most verbs introduce events. However, as we can see from our example this is not always the case. Events can be introduced by noun phrases, as in the move and the increase, and some verbs fail to introduce events, as in the phrasal verb took effect, which refers to when the event began rather than to the event itself. Similarly, light verbs such as make, take, and have often fail to denote events. In these cases, the verb is simply providing a syntactic structure for the arguments to an event expressed by the direct object as in took a flight.
Both rule-based and statistical machine learning approaches have been applied to the problem of event detection. Both approaches make use of surface information such as parts of speech information, presence of particular lexical items, and verb tense information. Fig. 22.23 illustrates the key features used in current event detection and classification systems.
Having detected both the events and the temporal expressions in a text, the next logical task is to use this information to fit the events into a complete timeline. Such a timeline would be useful for applications such as question answering
| Feature | Explanation |
| Character affixes Nominalization suffix Part of speech Light verb Subject syntactic category Morphological stem Verb root Wordnet hypernyms | Character-level prefixes and suffixes of target word\nCharacter level suffixes for nominalizations (eg. -tion)\nPart of speech of the target word\nBinary feature indicating that the target is governed by a light verb\nSyntactic category of the subject of the sentence\nStemmed version of the target word\nRoot form of the verb basis for a nominalization\nHypernym set for the target |
| Figure 22.23 Features commonly used in both rule-based and statistical approaches to event detection. | |
and summarization. This ambitious task is the subject of considerable current research but is beyond the capabilities of current systems.
A somewhat simpler, but still useful, task is to impose a partial ordering on the events and temporal expressions mentioned in a text. Such an ordering can provide many of the same benefits as a true timeline. An example of such a partial ordering would be to determine that the fare increase by American Airlines came after the fare increase by United in our sample text. Determining such an ordering can be viewed as a binary relation detection and classification task similar to those described earlier in Sec. 22.2.
Current approaches to this problem attempt to identify a subset of Allen's 13 temporal relations discussed earlier in Ch. 17, and shown here in Fig. 22.24. Recent evaluation efforts have focused on detecting the before, after and during relations among the temporal expressions, document date and event mentions in a text (Verhagen et al., 2007). Most of the top-performing systems employ statistical classifiers, of the kind discussed earlier in Sec. 22.2, trained on the TimeBank corpus (Pustejovsky et al., 2003b).