China
Africa
Explore chapters and articles related to this topic
Information Retrieval and Semantic Search
Published in Anuradha D. Thakare, Shilpa Laddha, Ambika Pawar, Hybrid Intelligent Systems for Information Retrieval, 2023
Anuradha D. Thakare, Shilpa Laddha, Ambika Pawar
The difficulty we tend to face within the technologically superior globe is the computer interprets the language or logic like a human. “Semantics” refers to the ideas or concepts specified by words, and semantic analysis is creating any topic (or search query) simple for a machine to interpret. Semantic mapping is concerned with visualizing relationships between concepts and entities (as well as relationships between connected concepts and entities). Semantic analysis needs policies to be outlined for the system. These policies are identical because of the manner we think of language and that we anticipate the computer to emulate. For instance, “ball is orange” could be a statement that a person can interpret that there’s one thing known as ball and it is orange in color, and therefore the human is aware of that orange suggests that color. However, for a computer, this is similar to an alien language. The concept of semantics here is that this sentence formation encompasses a structure in it. Subject—predicate-object or briefly type s-p-o. Wherever “ball” is subject, “is” is predicate, and “orange” is object. Likewise, there are different linguistic refinements that are utilized in the semantics analysis.
Hierarchical Bayesian model for the transfer of knowledge on spatial concepts based on multimodal information
Published in Advanced Robotics, 2022
Yoshinobu Hagiwara, Keishiro Taguchi, Satoshi Ishibushi, Akira Taniguchi, Tadahiro Taniguchi
Semantic mapping refers to methods of assigning meanings, such as vocabulary representing places or classes of objects and places, to an environment map held by the robot. A semantic map is a map to which such meanings are assigned. There are various approaches to semantic mapping [5]. In early studies on semantic mapping, methods of attaching object labels obtained by deep learning algorithms to an occupancy grid map as semantic attributes were proposed [22–25]. These studies enabled a robot to use the semantic information on objects in an occupancy grid map but not to use place information such as location names associated with spatial regions (e.g. kitchens) Studies on semantic mapping based on deep learning algorithms with labeled datasets for places have been conducted [6,7,26–31]. Sunderhauf et al. proposed a place-recognition method by applying a CNN to images obtained by a robot moving in an environment and assigning the obtained place classes to an occupied grid map [6]. A large dataset called the Place205 dataset was used to train the CNN. Pal et al. proposed DEDUCE [7], which combined a place-recognition CNN called Place365 [11] and an object-recognition CNN called You Only Look Once (YOLO) [32]. Experimental results indicated that this integrated approach provided a higher recognition accuracy than existing place-recognition methods.