Explanation techniques such as local interpretable model-agnostic explanation (LIME) provide reasons behind decisions made by machine-learning models. These methods typically use a set of features and their values as inputs and identify those that significantly influence the final decision. However, machine-learning models for entity matching operate on two sets of tokens or records, each representing an entity, to determine whether they refer to the same real-world entity. Explanations for entity-matching decisions are more convincing when they highlight contributing pairs of tokens within the pair of records, rather than focusing on individual tokens alone. In this sense, existing explanation techniques are insufficient for entity matching. Therefore, we propose a new method, Pair INterpretation for Entity matching (PINE), which takes two records as input, and outputs correlated token pairs as an explanation for an entity-matching decision. Our extensive experiments on public datasets demonstrate that the extracted token pairs exhibit strong correlations and serve as interpretable evidence for matching records.

To build up spontaneous conversation, it is important to select topics without a feeling of strangeness. When someone notices others are not interested in a topic, he/she tries to find a new topic. Then, he/she thinks of viewpoints of the conversation and selects a topic associated with the current topic from the viewpoints. To automate viewpoint-based topic selection, we present 5W viewpoint associative topic search. The method estimates the weights of 5W viewpoints (who, what, where, when and why) from conversation, to use an appropriate similarity to search for the next topic.

We describe the implementation of the proposed single user type CDMA adaptive interference canceller (AIC) with RAKE structure in a base station testbed, and evaluate its performance in the multiuser and multipath fading environment. Laboratory experiments demonstrate that the AIC receiver is much more near-far resistant than a conventional matched filter (MF) receiver in the multiuser case. When the power of the other users is 6 dB larger than that of the desired user, the AIC receiver can achieve a BER of 10/sup -3/ at C/PG=33.3% (eight users) in the 2-path fading channel, while the MF receiver cannot achieve the BER at C/PG of more than 20.8% (five users), it can achieve the BER at C/PG=33.3% in the equal power case. Furthermore, we evaluate the effect of transmission power reduction in a mobile station with transmission power control (TPC). Experimental results show that the required transmission power can be greatly reduced by 3.0 dB and 9.2 dB with the AIC receiver at C/PG=29.2% (seven users) and 33.3% (eight users), respectively.