Retrieval Effectiveness of Google on Reverse Image Search

This study investigates the reverse image search performance of Google, in terms of Average Precisions (APs) and Average Normalized Recalls (ANRs) at various cut-off points,on finding out similar images by using fresh Image Queries (IQs) from the five categories “Fashion,” “Computer,” “Home,” “Sports,” and “Toys.” The aim is to have an insight about retrieval effectiveness of Google on reverse image search and then motivate researchers and inform users. Five fresh IQs with different main concepts were created for each of the five categories. These 25 IQs were run on the search engine, and for each, the first 100 images retrieved were evaluated with binary relevance judgment. At the cut-off points 20, 40, 60, 80, and 100, both APs and ANRs were calculated for each category and for all 25 IQs. The AP range is from 41.60% (Toys—cut-off point 100) to 71% (Home—cut-off point 20). The ANR range is from 47.21% (Toys—cut-off point 20) to 71.31% (Computer—cut-off point 100). If the categories are ignored; when more images were evaluated, the performance of displaying relevant images in higher ranks increased, whereas the performance of retrieving relevant images decreased. It seems that the information retrieval effectiveness of Google on reverse image search needs to be improved.

A Comparison of Retrieval Result Relevance Judgments Between American and Chinese Users

Relevance judgment plays an extremely significant role in information retrieval. This study investigates the differences between American users and Chinese users in relevance judgment during the information retrieval process. 384 sets of relevance scores with 50 scores in each set were collected from 16 American users and 16 Chinese users as they judged retrieval records from two major search engines based on 24 predefined search tasks from 4 domain categories. Statistical analyses reveal that there are significant differences between American assessors and Chinese assessors in relevance judgments. Significant gender differences also appear within both the American and the Chinese assessor groups. The study also revealed significant interactions among cultures, genders, and subject categories. These findings can enhance the understanding of cultural impact on information retrieval and can assist in the design of effective cross-language information retrieval systems.

Expert Judgment Z-Numbers as a Ranking Indicator for Hierarchical Fuzzy Logic System

<p>In this article, the researchers main contribution is to investigate three factors which may correlate in implementation of Expert Judgment Z-Numbers as new Fuzzy Logic Ranking Indicator such as: expert relevance judgment or score, the expert confidence and the level of expertise. The Expert Judgment Z-Numbers then will be an input to the Hierarchical Fuzzy Logic System of Domain Specific Text Retrieval, along with other indicators such as Ontology BM25 Score, Fabrication Rate, Shia Rate and Positive Rate of hadith document. The results showed, the proposed system, with the additional new indicator of Expert Judgment Z-Numbers, may improve the original BM25 ranking function, by yielding better results on 26 queries, on all evaluation metrics that are measured in this research such as P@10, %no measures and MAP, and has achieved better results in 28 queries on P@10 alone, compared to the BM25 original score, that only yield better results in 2 queries on all evaluation metrics, and also yield better results in 4 queries on the MAP alone. The results proved that the proposed system has a capability to utilize the expert confidence and their relevant judgment that are represented in Z-Number, as an indicator to optimize the existing ranking function system and has a potential for a further research to be conducted on these domains. For the future works, the researchers would like to enhance this research by including a variety of expert’s level confidence and their judgment, also a new calculation to represent the value of Z-Numbers.</p>

Development of Indonesian Stemming Algorithms through Modification of Grouping, Sequencing and Removing of Affixes Based on Morphophonemic

Text documents stored on the system in an unstructured form, so that the information inside cannot be extracted directly. To be able to extract it, it takes text processing which is first carried out initial processing (preprocessing text) to convert text documents into more structured by selecting words that used as indexes. The smaller the index value, the more text documents are recognized on the system and the information is more easily extracted. The size of the index determined by the number of groups of words formed. To avoid forming many groups of words, then each word is changed to become a basic word first before grouping. The process of changing of affix word into a basic word using certain rules is called stemming. This research aims to produce a new Indonesian stemming algorithm named UG18 Stemmer algorithm, which can reduce or eliminate stemming errors such as over-stemming and under-stemming on existing stemming algorithms including the Enhanced Confix Stripping (ECS) Stemmer algorithm and the New Enhanced Confix Stripping (NECS) stemming algorithm. The method used is the morphophonemic process approach, which sees affixes as bound morphemes that experience phoneme changes, phoneme addition, and phoneme removal. The three processes are mapped, and Finite State Automata was made to obtain new affixed groups, sequences and new deletion methods that form the basis of the development of the UG18 Stemmer algorithm. This algorithm developed not using a list of decapitation rules used in pre-existing algorithms. Decapitation rules replaced with morphophonemic based elimination rules. Based on the evaluation results and testing of the UG18 Stemmer algorithm, it has a lower error rate compared to the results of stemming using NESC Stemmer. The result can be seen from the randomized test of 2500 word using Relevance Judgment validated by Indonesian language experts, from 1.48% over-stemming and 16.69% under-stemming using the NECS stemmer algorithm down to 0.12% overstemming and 0% understemming using the UG18 algorithm stemmer. Also, the UG18 Stemmer algorithm can improve the speed performance process in the information retrieval-based document similarity measurement application of 45.47% compared to using the ECS stemmer algorithm.