Chunking Strategy Research Articles

Enhancing RAG Performance Through Chunking and Text Splitting Techniques

In the world of Generative Artificial Intelligence (GenAI) and Large Language Models (LLM), Retrieval-Augmented Generation (RAG) has transformed the way we interact with data. Using RAG, these models can leverage new data contexts to respond to user queries and gain valuable insights. Behind the outstanding capabilities of RAG, a fundamental pre-processing step is present known as chunking. This step plays a crucial role in the effectiveness of these RAG-enhanced models. Chunking involves the breaking down of large text or documents into smaller segments of a fixed size. This allows the retriever to focus on smaller units at a time, making it easier to process and analyse the text. Finding the ideal chunking strategy can be a challenging task. Experimenting and analysis play a decisive role here, as different chunking strategies cater to different use cases. This paper, mainly targeted for an audience that is exploring RAG tuning techniques for higher accuracy, explores the various chunking techniques and their practical implementation using code snippets. After analysing the results for various use cases, the paper also suggests the best use cases for the different chunking strategies. Finally, it concludes by discussing the future potential and extending scope of RAG-enhanced applications.

Assessing individual differences in chunking strategy in visual working memory

Assessing individual differences in chunking strategy in visual working memory

Chunking Strategy to Improve Students' Reading Comprehension: An Action Research

The study focused on Chunking Strategy to Improve Student’s Reading Comprehension in the Nonequivalent Control Group Design, a form of quasi-experimental group design. This research consisted of two groups (experimental and controlled groups). Both groups were given pre-test and post-test to determine whether the chunking strategy could improve the students' reading comprehension. This research was conducted on the Junior High School Students of Sibugay Integrated School. Moreover, The researchers used the Reading Comprehension Test Research Instrument by Wigfield and Guthrie (1995) which was developed to evaluate many facets of students' reading motivation. The findings revealed a significant difference in the post-test and pretest scores of the students who underwent the chunking strategy. The findings of this study affirmed that using a chunking method was useful for teaching reading comprehension in the educational process. The pre-test and post-test scores of the students showed significant improvement. By using this strategy, teachers may make it easier for students to comprehend and retain material, which will boost their performance.

The Use of Chunking Strategy in Teaching Reading Comprehension

This research was aimed to find out whether using chunking strategy was effective in teaching reading comprehension. In this case, the researcher used quantitative data approach by using experimental design. There were two classes at ninth grade of SMPN 1 Nisam North Aceh as the sample of this research. There were class IX/3 as experimental consist of 24 students and IX/2 as control group consist of 22 students. In experimental group, the researcher used chunking strategy in teaching reading comprehension. After teaching learning process for three meetings, the researcher gave post-test. The result was shown that the students’ achievement in experimental group was higher than control group. It was proved by the average score from pre-test of experimental group was 66,8 and control group was 64,5. Furthermore, the result of post-test of experimental group was 83,2 and control group was 70,3. It could be concluded that using chunking strategy was effective in teaching reading comprehension.

Fine-grained-based multi-feature fusion for occluded person re-identification

Many previous occluded person re-identification(re-ID) methods try to use additional clues (pose estimation or semantic parsing models) to focus on non-occluded regions. However, these methods extremely rely on the performance of additional clues and often capture pedestrian features by designing complex modules. In this work, we propose a simple Fine-Grained Multi-Feature Fusion Network (FGMFN) to extract discriminative features, which is a dual-branch structure consisting of global feature branch and partial feature branch. Firstly, we utilize a chunking strategy to extract multi-granularity features to make the pedestrian information contained in it more comprehensive. Secondly, a spatial transformer network is introduced to localize the pedestrian’s upper body, and then introduce a relation-aware attention module to explore the fine-grained information. Finally, we fuse the features obtained from the two branches to obtain a more robust pedestrian representation. Extensive experiments verify the effectiveness of our method under the occlusion scenario.

Working Memory for Spatial Sequences: Developmental and Evolutionary Factors in Encoding Ordinal and Relational Structures.

Sequence learning is a ubiquitous facet of human and animal cognition. Here, using a common sequence reproduction task, we investigated whether and how the ordinal and relational structures linking consecutive elements are acquired by human adults, children, and macaque monkeys. While children and monkeys exhibited significantly lower precision than adults for spatial location and temporal order information, only monkeys appeared to exceedingly focus on the first item. Most importantly, only humans, regardless of age, spontaneously extracted the spatial relations between consecutive items and used a chunking strategy to compress sequences in working memory. Monkeys did not detect such relational structures, even after extensive training. Monkey behavior was captured by a conjunctive coding model, whereas a chunk-based conjunctive model explained more variance in humans. These age- and species-related differences are indicative of developmental and evolutionary mechanisms of sequence encoding and may provide novel insights into the uniquely human cognitive capacities.SIGNIFICANCE STATEMENT Sequence learning, the ability to encode the order of discrete elements and their relationships presented within a sequence, is a ubiquitous facet of cognition among humans and animals. By exploring sequence-processing abilities at different human developmental stages and in nonhuman primates, we found that only humans, regardless of age, spontaneously extracted the spatial relations between consecutive items and used an internal language to compress sequences in working memory. The findings provided insights into understanding the origins of sequence capabilities in humans and how they evolve through development to identify the unique aspects of human cognitive capacity, which includes the comprehension, learning, and production of sequences, and perhaps, above all, language processing.

STRATEGIES OF PATTERN GENERALIZATION FOR ENHANCING STUDENTS’ ALGEBRAIC THINKING

Mathematics is seen as a science of pattern. Identifying and using patterns is the essence of mathematical thinking for children to improve algebraic thinking from their early schooling. The pattern is an arrangement of objects that have regularities or properties that can be generalized. Therefore, it is essential to know the strategies used by students in generalizing patterns and how students think in these processes. This study is descriptive research with a mixed quantitative-qualitative approach that aimed to investigate student’s algebraic thinking using various strategies to generalize the visual pattern. An instrument about the linear geometric growing pattern was administrated to 75 upper primary school students (grades 5-6) and 81 lower secondary students (grades 7-8) in two private schools in Semarang, Indonesia. The results showed that students used different pattern generalization strategies. The student generally preferred recursive, chunking, and functional approaches in each generalization task, whereas few used counting from drawing strategies to generalize patterns. The use of the recursive strategy decreased, whereas the chunking strategy and the functional strategy increased across grades 5-8 for the problems. The results also showed the student who used the recursive and chunking strategy preferred to change visual patterns into rows of numbers. Hence, they adopt a numeric approach by finding the common difference of visible pattern in each step.

THE EFFECTIVENESS OF CHUNKING STRATEGY IN DEVELOPING STUDENT'S ABILITY IN STORY RETELLING

Abstract: This research aims at finding out whether or not the use of chunking strategy is effective in developing students’ ability of story retelling to the tenth-grade students at MAN 01 Pekalongan. Quasi experimental research was applied as the method with two groups : experimental and control group. The total subject of this research was 52 students which were taken from two classes of X IPS 1 as a control group and X AGAMA 2 as an experimental group. Face and content validity were done by asking the English teacher and advisor to be expert judgements to validate the instrument by using a rubric appearance of test for face validity and its conformity by curriculum as a content validity. Inter-rater reliability in giving scores to the students’ was done to measure the reliability of the test. This research used SPSS in analysing the data of speaking test through several activities. They were pre-test before treatments, and post-test after treatments. The result of the research shows the significant difference in the story retelling ability between tenth grade students who were taught by using chunking strategy and those were not taught by using it. It can be seen in the result of mean score in the post-test of experimental class which is 89 and control class which is 83 and as a result Sig. (2-tailed) is 0.044 0.050. Therefore, the null hyothesis of this study is rejected. It means that the use of chunking strategy had significantly improved the tenth-grade students story retelling ability in English teaching and learning process of the experimental class at MAN 01 Pekalongan in the academic year 2019/2020. Keywords: chunking, teaching speaking strategy, story retelling

Chunking Strategy; in Enhancing Fourth Semester Students’ Reading Skill at English Department of Muhammadiyah University

Teaching reading skill to the students needs applying the appropriate strategy to stimulate students’ understanding to meet the main purpose in the teaching and learning process. It was expected that the strategy itself should be able to be used by the teacher to design the better ways for the students in order to acquire important information. It is realized that many students now need as much information as possible in comprehending all the texts they read. To overcome this problem, the researcher used chunking strategy to improve students’ reading skill. The researcher used true experiment to design this research, then the pre-test was used to find out whether this group in the same level and the post-test was used to detect the progress of the students. The total numbers of the students who involved in the pre-test and post-test were 20. The data showed that the information in the post-test that; mean score was 70.50, and the pre-test got score was 61.85 then the standard deviation in the post-test was 10.32 and the standard deviation in pre-test is 8.27. It could be concluded that the students get significant progress in their understanding of reading skill by using Chunking strategy in English Department.

学习材料组块方式对相似词长时记忆的影响

学习材料组块方式对相似词长时记忆的影响

Production of Supra-regular Spatial Sequences by Macaque Monkeys

Understanding and producing embedded sequences in language, music, or mathematics, is a central characteristic of our species. These domains are hypothesized to involve a human-specific competence for supra-regular grammars, which can generate embedded sequences that go beyond the regular sequences engendered by finite-state automata. However, is this capacity truly unique to humans? Using a production task, we show that macaque monkeys can be trained to produce time-symmetrical embedded spatial sequences whose formal description requires supra-regular grammars or, equivalently, a push-down stack automaton. Monkeys spontaneously generalized the learned grammar to novel sequences, including longer ones, and could generate hierarchical sequences formed by an embedding of two levels of abstract rules. Compared to monkeys, however, preschool children learned the grammars much faster using a chunking strategy. While supra-regular grammars are accessible to nonhuman primates through extensive training, human uniqueness may lie in the speed and learning strategy with which they are acquired.

Simple and Complex Working Memory Tasks Allow Similar Benefits of Information Compression.

Complex working memory span tasks were designed to engage multiple aspects of working memory and impose interleaved processing demands that limit the use of mnemonic strategies, such as chunking. Consequently, the average span is usually lower (4 ± 1 items) than in simple span tasks (7 ± 2 items). One possible reason for the higher span of simple span tasks is that participants can take advantage of the spare time to chunk multiple items together to form fewer independent units, approximating 4 ± 1 chunks. It follows that the respective spans of these two types of tasks could be equal (at around 4 ± 1) if stimulus lists exclusively used nonchunkable stimulus items. To manipulate the chunkability of the stimulus lists, our method involved a measure of their compressibility, i.e., the extent to which a pattern exists that can be detected and used as a basis of chunk formation. We predicted an interaction between the types of tasks and chunkability/compressibility, supporting a single higher span for the condition in which a simple span task was combined with chunkable items. The three other conditions were predicted to prevent chunking processes, either because the interleaved processing task did not allow any chunking process to occur or because the noncompressible material inherently limited the chunkability of information. The prediction that chunking is important solely in simple spans was not confirmed: Effects of information compression contributed to performance levels to a similar extent in both tasks according to a theoretically-based metric. This result suggests that i) complex span tasks might overestimate storage capacity in general, and ii) the difference between simple and complex span performance levels must rest in some mechanism other than prevention of a chunking strategy by the interleaved processing task in complex span tasks.

Differences in chunking behavior between young and older adults diminish with extended practice

Previous research found reduced motor chunking behavior in older adults compared to young adults. However, it remains unclear whether older adults are unable to use a chunking strategy or whether they are just slower in developing them. Our goal was to investigate the effect of extended practice on the development of chunking behavior in healthy older adults. A group of young and a group of healthy older adults between 74 and 85 years of age visited the lab on 2 days. A sequence of 3 and a sequence of 6 elements were both practiced 432 times in a discrete sequence production task. We found that age differences in chunking behavior, as measured by the difference between initiation and execution of the sequence, diminish with extended practice. Furthermore, in older, but not in young adults, slow responses that are often interpreted as the first response of a next motor chunk were associated with a finger that was also slow during performance of the random sequences. This finding calls for more attention to biomechanical factors in future theory about aging and sequence learning.

Caching for where and what: evidence for a mnemonic strategy in a scatter-hoarder.

Scatter-hoarding animals face the task of maximizing retrieval of their scattered food caches while minimizing loss to pilferers. This demand should select for mnemonics, such as chunking, i.e. a hierarchical cognitive representation that is known to improve recall. Spatial chunking, where caches with the same type of content are related to each other in physical location and memory, would be one such mechanism. Here we tested the hypothesis that scatter-hoarding eastern fox squirrels (Sciurus niger) are organizing their caches in spatial patterns consistent with a chunking strategy. We presented 45 individual wild fox squirrels with a series of 16 nuts of four different species, either in runs of four of the same species or 16 nuts offered in a pseudorandom order. Squirrels either collected each nut from a different location or collected all nuts from a single location; we then mapped their subsequent cache distributions using GPS. The chunking hypothesis predicted that squirrels would spatially organize caches by nut species, regardless of presentation order. Our results instead demonstrated that squirrels spatially chunked their caches by nut species but only when caching food that was foraged from a single location. This first demonstration of spatial chunking in a scatter hoarder underscores the cognitive demand of scatter hoarding.

The effects of task-relevant saccadic eye movements performed during the encoding of a serial sequence on visuospatial memory performance.

Visuospatial working memory (VSWM) is a set of cognitive processes used to encode, maintain and manipulate spatial information. One important feature of VSWM is that it has a limited capacity such that only few items can be actively stored and manipulated simultaneously. Given the limited capacity, it is important to determine the conditions that affect memory performance as this will improve our understanding of the architecture and function of VSWM. Previous studies have shown that VSWM is disrupted when task-irrelevant eye movements are performed during the maintenance phase; however, relatively fewer studies examined the role of eye movements performed during the encoding phase. On one hand, performing eye movements during the encoding phase could result in a stronger memory trace because the memory formation is reinforced by the activation of the motor system. On the other hand, performing eye movements to each target could disrupt the configural processing of the spatial array because the spatial representation has to be updated with each movement to maintain perceptual stability. Therefore, this work was conducted to examine whether task-relevant saccadic eye movements performed during the encoding phase of a visuospatial working memory task affect the recall of serially presented targets. Results from two experiments showed that average recall accuracy was significantly higher when the spatial array (set size ≥ 7) was encoded using a covert strategy-that is, while participants fixated on a central target, in comparison to an overt strategy-that is, while participants moved their eyes to fixate on each target. Furthermore, the improvement in accuracy was evident only for targets presented in the first half of the sequence, suggesting that the primacy effect is modulated by the presence of eye movements. We propose that executing saccades during encoding could interfere with the ability to use a chunking strategy or disrupt active visualization of the configuration. In conclusion, this is the first study to show that task-relevant saccadic eye movements performed during encoding may actually reduce the spatial span of VSWM. These results extend the current knowledge about the role of eye movements in VSWM, and have implications for future studies investigating the VSWM.

The Effects of Sentence Repetition-Based Working Memory Treatment on Sentence Comprehension Abilities in Individuals With Aphasia.

This study investigated whether sentence repetition-based working-memory (SR-WM) treatment increased sentence-repetition abilities and the treatment effects generalized to sentence-comprehension abilities, WM-span tasks, and general language-assessment tasks. Six individuals with aphasia participated in the study. The treatment consisted of 12 sessions of approximately 1 hr per day, 3 times per week. The SR-WM treatment protocol followed components including maintenance and computation of linguistic units by facilitating a chunking strategy. We manipulated the length and syntactic structures of the sentence-repetition stimuli using a limited set of vocabulary. Participants demonstrated significant increased repetition ability in treated and untreated sentences after treatment. Furthermore, they showed generalization effects on the sentence-comprehension task, WM measures, and general language tasks, but with some differential patterns, depending on task demands. The SR-WM treatment approach, by manipulating syntactic structures and minimizing top-down semantic processing, elicited increased performance on sentence repetition as well as other linguistic domains. Results indicated that it is clinically and theoretically important to examine whether WM treatment serves as a potentially underlying treatment approach that facilitates the distributed network associated with language processing.

Disruption of Broca's Area Alters Higher-order Chunking Processing during Perceptual Sequence Learning.

Because Broca's area is known to be involved in many cognitive functions, including language, music, and action processing, several attempts have been made to propose a unifying theory of its role that emphasizes a possible contribution to syntactic processing. Recently, we have postulated that Broca's area might be involved in higher-order chunk processing during implicit learning of a motor sequence. Chunking is an information-processing mechanism that consists of grouping consecutive items in a sequence and is likely to be involved in all of the aforementioned cognitive processes. Demonstrating a contribution of Broca's area to chunking during the learning of a nonmotor sequence that does not involve language could shed new light on its function. To address this issue, we used offline MRI-guided TMS in healthy volunteers to disrupt the activity of either the posterior part of Broca's area (left Brodmann's area [BA] 44) or a control site just before participants learned a perceptual sequence structured in distinct hierarchical levels. We found that disruption of the left BA 44 increased the processing time of stimuli representing the boundaries of higher-order chunks and modified the chunking strategy. The current results highlight the possible role of the left BA 44 in building up effector-independent representations of higher-order events in structured sequences. This might clarify the contribution of Broca's area in processing hierarchical structures, a key mechanism in many cognitive functions, such as language and composite actions.

Chunking improves symbolic sequence processing and relies on working memory gating mechanisms.

Chunking, namely the grouping of sequence elements in clusters, is ubiquitous during sequence processing, but its impact on performance remains debated. Here, we found that participants who adopted a consistent chunking strategy during symbolic sequence learning showed a greater improvement of their performance and a larger decrease in cognitive workload over time. Stronger reliance on chunking was also associated with higher scores in a WM updating task, suggesting the contribution of WM gating mechanisms to sequence chunking. Altogether, these results indicate that chunking is a cost-saving strategy that enhances effectiveness of symbolic sequence learning.

THE EFFECTIVENESS OF USING CHUNKING STRATEGY TO IMPROVE STUDENTS’ READING COMPREHENSION AT THE SECOND YEAR OF SMP NEGERI 2 BAROMBONG

The objective of this research was to know whether chunking strategy effective to improve students’ reading comprehension of the second year of SMP Negeri 2 Barombong and found the students’ attitude in learning using chunking strategy. The research design applied non-equivalent control group design which is one of the forms of quasi experimental design consisting of two groups (Experimental group and controlled group) each group covering 6 meetings. This research focused on improving the students’ reading comprehension in comprehending narrative text. The instruments used were reading text, observation sheet, and questionnaire.Further, the data analysis techniqueused in this research was statisticalanalysis.The findings of this research showed that chunking strategy effective to improve the students reading comprehension in which the mean score of students’ post-test of experimental group was 56. It was higher than the controlled group that was only got 23.This research was successful to solve the students’ problem in comprehending thetextby using chunking strategy.

Contextual effects in visual working memory reveal hierarchically structured memory representations.

Influential slot and resource models of visual working memory make the assumption that items are stored in memory as independent units, and that there are no interactions between them. Consequently, these models predict that the number of items to be remembered (the set size) is the primary determinant of working memory performance, and therefore these models quantify memory capacity in terms of the number and quality of individual items that can be stored. Here we demonstrate that there is substantial variance in display difficulty within a single set size, suggesting that limits based on the number of individual items alone cannot explain working memory storage. We asked hundreds of participants to remember the same sets of displays, and discovered that participants were highly consistent in terms of which items and displays were hardest or easiest to remember. Although a simple grouping or chunking strategy could not explain this individual-display variability, a model with multiple, interacting levels of representation could explain some of the display-by-display differences. Specifically, a model that includes a hierarchical representation of items plus the mean and variance of sets of the colors on the display successfully accounts for some of the variability across displays. We conclude that working memory representations are composed only in part of individual, independent object representations, and that a major factor in how many items are remembered on a particular display is interitem representations such as perceptual grouping, ensemble, and texture representations.