An Investigation into the Utility of the Probabilistic Interpretation to the Hyper Analogue to Language (hal) Space

In Song and Bruza [15], they introduce a framework for Information Retrieval(IR) based on Gardenfor’s three tiered cognitive model; Conceptual Spaces[5]. They instantiate the conceptual space to generate informational inferences via the high dimensional conceptual space which are later used for ad-hoc retrieval. The implementation utilizes the Hyper Analogue to Language (HAL) algorithm [10] to build such a high dimensional conceptual space and generate inferences by employing Barwise and Seligman’s theory of Information Flow[2]. In this report, we propose an alternative implementation of the conceptual space by using a probabilistic HAL (pHAL) space as the basis for the conceptual space. To evaluate whether converting to such an implementation is beneficial we have performed an initial investigation comparing HAL and pHAL using the concept combination process for query expansion. This was performed across a range of parameters that are involved in the construction of the (p)HAL space and combination techniques. The influence on the retrieval effectiveness when applying such methods to query expansion was the basis of comparison. Our results indicated that pHAL is a competitive alternative to the original HAL method. Please note that this is a preliminary investigation which serves only to identify whether further analysis is warranted. The results reported in this study while not conclusive serve as an indication of the potential of a probabilistic alternative.