ROLE OF FRAME STRUCTURE IN
THE
DEVELOPMENT OF KRS FOR LEARNING
DIALOGUES
Harjit Singh
Assitant Professor, Department of Linguistics and CSTLs, Indira Gandhi National Tribal University,
Amarkantak, M.P. (484887), (India).
E-mail: harjitsingh.jnu@gmail.com
Reception: 22/11/2022 Acceptance: 07/12/2022 Publication: 29/12/2022
Suggested citation:
Harjit Singh. (2022). Role of frame structure in the development of KRS for learning dialogues. 3C Empresa.
Investigación y pensamiento crítico, 11(2), 250-261. https://doi.org/10.17993/3cemp.2022.110250.250-261
https://doi.org/10.17993/3cemp.2022.110250.250-261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
250
ABSTRACT
Dialogues are building blocks of tasks and non-tasks of communication, which happen between
objects in the universe. Each dialogue is a source of linguistic knowledge within a natural language
that explains and elaborates with frame structure in general. In this paper, it is noticed that various
forms like (nouns, pronouns, yes-no questions and deletion) are essential part of each dialogue
structure (DSS) in Chandan’s work 
/Roots. With the help of frames, knowledge representation
system (KRS) is prepared for such dialogues in Punjabi. On the other hand, it is argued that highest
numbers of nouns are total 45 in DS2 and only 1 deletion case finds in DS3. While DS1 and DS2 both
have similar number of 2-2 cases of yes-no questions. The overall evaluation is successfully matched
with proposed an algorithm based on frames.
KEYWORDS
/Roots, KRS, Frames, Nouns, Pronouns, Deletion.
https://doi.org/10.17993/3cemp.2022.110250.250-261
251
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
1. INTRODUCTION
The universe consists of number of objects and each one has not a particular form but also certain
characteristics that are source of information and knowledge. In this context, it is assumed that natural
language is a very complex object and it has many layers and levels of knowledge representation. By
introducing frames, it means that they are best ways to serve such knowledge. In general, frames
report stereotyped situations and are essential part of frame system (Minsky, 1974). They are complete
package of information for an object/a concept in spoken and written discourse. They look like an
individual type; an abstract type and also were prototypes and exemplars (Steels, 1978).
On the other hand, when an object comes with huge numbers and classes then it is called generic
frame (Brachman and Levesque, 2004). Indeed, it is an important to structure information and
knowledge with the help of mental models, semantic networks, scripts, plans and frames (Crowley,
2012). Based on Minsky’s frames and Fillmore’s frames and Schank’s scripts, a probablisitic model
has already been designed to define corpus related matters and coherent has also been increased
(Ferrarro and Durme, 2016). Also, it is noted down that frames oriented knowledge systems are
functioning well in Google and Siri like platforms (Boroujerdi, 2018). At the textual level, it is also
seen that knowledge is widely depended upon the context so that contextual study is as eqully
important. When the context is explored then it is useful for terminology in the text. For this purpose,
frames help to understand terminology within one word to group of words and group of words to
another higher category in the text (Faber and Cabezas-Garcia, 2019). Likewise text, spoken
conversation is a set of dialogues, which sometimes consist of group of four, five words and
sometimes more. But today, the dialogues are going to be systemized with frames for special tasks
whether it belongs to a doctor who tries to manage bad news with a patient and in this way, both they
share same information (Blache and Houles, 2021).
In this paper, it is tried to analysis yes-no questions, noun-to-pronoun shift and deletion like few cases
in Punjabi with frames and also present knowledge representation system (KRS). There are total six
sections. First section discusses frames and KRS for Punjabi. Second section focuses on historical
studies of frame structure and recent works. Third section indicates aims and objectives. Fourth
section describes methodology (type of data sets and arrangement). Fifth section presents results and
shows an algorithm. Last section draws conclusion and gives direction to future work. In brief,
knowledge representation system (KRS) for Chandan’s work /Roots (2006) is shown in Fig. 1.
https://doi.org/10.17993/3cemp.2022.110250.250-261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
252
Fig. 1. KRS for dialogue structure.
Above fig. 1 shows that there are four dialogue structures where nominal contains /Roots, /
earth, /poem, /mother, /paper, /writer, /colour etc in DS1. Second DS2 has /he,
/his/her like third personal pronouns and reflexives respectively. Whereas, it seems that /yes and
8/no one word yes-no questions notice in DS3. Similarly, personal pronouns such as /he, and /
his/her like reflexives apppear in ommitted form in DS3.
2. RELATED WORKS
Frames depict existed and flowing knowledge into dialogues and discourses (Thagard, 1984).
Dialogues provide wonderful platform to discuss events, situations and tasks/non-tasks. Emotions
within dialogues are captured by interface techniques (e.g. two tier mechanism) as suggested by
(Ruttkay and Pelachaud, 2005). Spoken dialogue systems are performing well when they introduce with
tasks/non-tasks (Jokinen and McTear, 2010).
On the other hand, it is said that the role of participants’ impact on turns taking and maintains
information flow within dialogue system (Thompson, 2013). Similarly, sets of phrases and small
utterances of any dialogue can also be analyzed with frames (Khan, 2013). Frames are easily discharing
knowledge through slots, values and so on. They are good source defining any particular doman in any
corner of the world (Nazaruks and Osis, 2017). ASR and n-gram features are another way to track
dialogue situations and they generalize dialogue contexts (Rudnicky et al., 2016). The use of ontology
controls both users and robot to model dialogues (D'Haro, 2019).
Few factors like (the choice of word order, pause and frequency) also show personalities of characters’
during dialogue processing (D'Haro, 2020). Based on dialogue or conversation between people or
https://doi.org/10.17993/3cemp.2022.110250.250-261
253
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
people with objects, it is necessary to capture frame knowledge to develop modern technology where
the route direction could be simple to improve the manufacturing work (Simonova and Kapitonova,
2019). The frame knowledge in the form of frame semantics has also been adopted to see the
relatedness between lexemes and to check them appropriately (Verdaguer, 2020).
It seems that frames also help to understand dialogues, particularly ‘inner dialogues’ where the special
focus is given upon speakers input and the mental state (Lopez-Soto, 2021). Regarding designing the
set of conversation within a dialogue of any language, it is an essential to use linguistic knowledge in
frames so that a dialogue can fully be represented in a systematic way (Chandrasegaran and Lioyd and
Akdag Salah, 2022). Followed by linguistics, it is realized that frames with their semantic knowledge,
they become effective tool to organize dialogues in English and German for the purpose of detecting
disasters for the society’s welfare. It has been argued that existing system “PAFIBERT” is again trained
to improve the accuracy and so on (Skachkova and Kruijff-Korbayova, 2021).
Based on semantics, “framenet Brasil” has been introduced to improve contextual domain and
generalizing sentences and texts computationally (Torrent et al., 2022). In this direction, “the research
group of Düsseldorf” has presented the history of frames in relation to linguistics and cognitive science.
It has been found that semantics and commmon sense knowledge are essential to develop linguistic
frame model that covers word claases in natural language (Löbner, 2021). Frame knowledge has also
been applied to discuss metaphors and it accurately mapping the metaphors in English (Stickles et al.,
2014).
3. AIMS AND OBJECTIVES
To survey dialogue and knowledge representation system.
To find out nominal, noun to pronoun shift, yes-no question and deletion like cases into dialogue
structures.
To analysis dialogue structures (DSs) through frames.
To present an algorithm based on KRS.
4. METHODOLOGY
It is declared that Chandan’s work /Roots is selected to discuss few cases of nominal, shifting and
etc. in dialogue structures (DSs). First, a single noun category is searched which is used to address,
order and request against total no of nouns. Then pronouns, yes-no questions and deletion are selected
one by one. In this procedure, it is suggested that each dialogue should be treated like a frame and it
would be subject to an algorithm.
5. RESULTS
There are three dialogues like (/Roots, /Instrument, and /Peacock) have been extracted from
Chandan’s work /Roots. In /Roots (DS1
) total 26 nouns and 11 pronouns are found whereas
noun-to-pronoun shift is commonly appeared. Only 02 cases find in yes-no questions. In /Instrument
(DS2), total 45 nouns and 05 pronouns are searched. Likewise /Roots (DS1), again 2 cases notice in
yes-no questions. Finally, in /Peacock (DS3), total 28 nouns and 02 pronouns (including 02 yes-no
questions and 02 deletion cases) are found. Fig. 2. shows complete analysis for DS1, DS2 and DS3.
https://doi.org/10.17993/3cemp.2022.110250.250-261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
254
Fig. 2. Total no of four variables in dialogue structures.
According to Fig. 2 it is shown that deletion does not find in DS1 and DS2 except total 2 in DS3. On the
other hand, yes-no questions are total 3 find in DvS1 and only 2-2 find in DS2 and DS3 respectively.
Here frame representation is significant to analysis above dialogue structures (DS1, 2, 3) one by one.
5.1. NOUNS IN FRAME STRUCTURE
Nouns indicate towards person, place and thing in the universe. In general, common, proper and mass
are kinds of nouns in natural language. Table 1 shows how frames are applied for nouns.
Table 1. Nouns in frames.
Nouns
Yes-No Questions
Deletion
2
2
2
28
0
2
5
45
0
2
26
DS1 DS2 DS3
https://doi.org/10.17993/3cemp.2022.110250.250-261
255
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
Slot Value Type
(Dhreja)
Sex Male Human
Age 36 yrs Biological
Home Place Stay
(God)
Sex Male Spiritual
Age Unimaginable Non-Biological
Place Everywhere Stay
Religion No Belief and Trust
(Fire)
Sex Male Deity
Element Cooking Life and Death bearer
(Well)
Sex Male Non-Human
Item Storage Big/Small Size
Place Somewhere Village
(Peacock)
Sex Male Non-Human
Age 20 yrs Biological
Home Rain forests Stay
(Workers)
Sex Male/Female Human
Age 40 yrs Biological
Home Place Stay
https://doi.org/10.17993/3cemp.2022.110250.250-261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
256
Table 1 shows that (Dhreja), (God), (Fire), (Well), and (Workers)
are common nouns. Each
example is a complete set of knowledge appearing with slot, value and type.
5.2. PRONOUNS IN FRAME STRUCTURE
Pronouns mostly stand against nouns in order to accomplish substitution tasks in natural language.
Apart from personal pronouns, they are also called reflexives, reciprocals, zero and so on. However,
only personal pronouns like (he), (it), (that) and (I) are found. The analysis for personal pronouns is
mentioned in Table 2.
Table 2. Pronouns in frames.
Table 2 shows that frames explain (he), (it), (that) and (I) like personal pronouns in a better way.
5.3 YES-NO QUESTIONS IN FRAME STRUCTURE
‘Polar questions’ and ‘general questions’ are selected for yes-no questions in linguistics. Each one receives
one word answer (either affirmative or negative). See Table 3.
Table 3. Yes-no Questions in Frames.
Slot Value Type
(He)
Sex Male/Female Human
Age 16 yrs Biological
Home Place Stay
(It)
Sex Male/Female Non-Human
Age Not countable Non-Biological
Use Need based Product (Pen, Book
etc)
(That)
Sex Male/Female Human/Non-Human
Age 32 yrs Biological
Home Place Stay
(I)
Sex Male/Female Human
Age 20 yrs Biological
Home Place Stay
Slot Value Type
https://doi.org/10.17993/3cemp.2022.110250.250-261
257
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
Table 3 demonstrates that /Yes as affirmative answer comes under slot which is filled up by sex, age
and query like variables. Value and type is another sort of information source for yes-no questions.
Value which means that a person may belong to male category or not and it may be 24 yrs old. Under
query, it is shown that satisfaction is kept against the slot and it falls down either accepted or rejected.
5.4 DELETION IN FRAME STRUCTURE
Deletion means to see omit items in spoken and written dialogues. For instance,  /Flying
Peacocks is deleted in the phrase ‘  ◌ …. -  but remember that this ____empty
space is filled up with ‘ ’only. See Table 4.
Table 4. Deletion in Frames.
Table 4 indicates that ‘ ’/Flying Peacock is not appeared in the second line because it is in
omitted mode.
5.5 PLANNING FOR AN ALGORITHM
It is pointed out that frames can generalize dialogue structures in a better way. Slot, value and type do
simplification for mentioned each dialogue structure and they provide a complete package of
information. In this regard, following algorithm is proposed for DS1, DS2, and DS3.
Step 1
Check nouns in dialogues (1, 2 and 3)
Collect all possible nouns
Select each one for frames
Step 2
Check pronouns in dialogues (1, 2 and 3)
/Yes
Sex Male/Female Human
Age 24 yrs Biological
Query Satisfaction Acceptable/Rejection
Slot Value Type
  (Flying
Peacock)
Sex Male Non- Human
Age 20 yrs Biological
Home Rain forest Stay
https://doi.org/10.17993/3cemp.2022.110250.250-261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
258
Collect all possible pronouns
Select each one for frames
Step 3
Check yes-no questions in dialogues (1, 2 and 3)
Collect all possible polar questions
Select each one for frames
Step 4
Check deletion in dialogues (1, 2 and 3)
Collect all possible deletion/omitted cases
Select each one for frames
It is represented in Fig. 3.
Fig 3. Steps of an Algorithm.
As per given an algorithm, it is suggested that there are total four steps where there is generally three
conditions are applied. In step 1, capital N denotes nouns, capital P denotes pronoun in step 2. Similarly,
it is seen that Y indicates (yes-no questions) and last D denotes deletion case in step 3 and 4
respectively. All four steps with corresponding N, P, Y, and D must follow the sequence of
(check>select>frame) at the execution time.
https://doi.org/10.17993/3cemp.2022.110250.250-261
259
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
In other words, it is simply mentioned that noun, pronoun, yes-no question and deletion kind of cases
are seen in dialogue 1,2,3. All they pass through check, select and giving frame slot three criteria. It is
pointed out that total number (as already given in fig 2) is successfully identified.
6. CONCLUSION AND FUTURE WORKS
Dialogues usually contain nouns, pronouns, anaphors and zero forms. Frames are used to explain them
in the work of 
/Roots. The highest number of nouns is 45 that find in DS1 whereas a single
deletion case is found in DS3. It is noticed that above mentioned algorithm fairly defines each DS and
in future, it could be modified to incorporate other dialogues like ◌   ’, , 
,    ◌ , ◌  , -◌,  ,   , ,  L
◌ ,     etc.
REFERENCES
[1]
Blache, P., Houlès, M. (2021). Common Ground, Frames and Slots for Comprehension in Dialogue
Systems. Proceedings of 7th International Conference on Natural Language Computing (NALT),
London, UK, 33-43.
[2] Boroujerdi, E. G. (2018).
Interactive Question Answering Using Frame-Based Knowledge
Representation. (Thesis, York University, Toronto).
[3] Brachman, R. J., Levesque, H. J. (2004). Knowledge Representation and Reasoning. Amsterdam:
Elsevier.
[4] Chandrasegaran, S., Lloyd, P., Akdag Salah, A. (2022). Identifying problem frames in design
conversation. Design Research Society: Biennial Conference Series, Bilbao, Spain, 1-16.
[5] Collins, A., & Bobrow, D. G. (Eds.). (2017).
Representation and Understanding: Studies in
Cognitive Science. New York: Academic Press.
[6] Crowley, J. L. (2012). Intelligent Systems: Reasoning and
Recognition. training, 1000, 1. Retrieved from http://crowley-coutaz.fr//jlc/Courses/2010/
ENSI2.SIRR/ENSI2.SIRR.S1.pdf.
[7] D'Haro, L. F., Banchs, R. E., & Li, H. (Eds.) (2019).
9th International Workshop on Spoken
Dialogue System Technology
. Lecture Notes in Electrical Engineering (vol. 579) Singapore:
Springer.
[8] D'Haro, L. F., Callejas, Z., and Nakamura, S. (Eds.) (2020). Conversational Dialogue Systems for
the Next Decade. Lecture Notes in Electrical Engineering (vol. 704) Singapore: Springer.
[9] Faber, P., Cabezas-García, M. (2019). Specialized Knowledge Representation: From Terms to
Frames, Research in Language, 17 (2), 197-211. doi: 10. 18778/1731-7533.17.2.06
[10] Ferraro, F., Durme, V. B. (2017). A Unified Bayesian Model of Scripts, Frames and Language.
Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence (AAAI-16), Phoenix,
Arizona, USA, 2601-2607.
[11] Jokinen, K., and McTear, M. (2010). Spoken Dialogue Systems
. Toronto: Morgan and Claypool
Publishers.
[12] Löbner, S. (2021). Frames at the Interface of Language and Cognition.
Annual Review of
Linguistics, (7), 261-284. doi: 10.1146/annurev-linguistics-042920-030620
[13] Minsky, M. (1974). A framework for representing knowledge/Marvin Minsky. In P. H. Winston
(Ed.), The psychology of computer vision. New York: McGraw-Hill Book.
[14] Nazaruks, V., and Osis, J. (2017). A Survey on Domain Knowledge Representation with Frames.
Proceedings of the 12th International Conference on Evaluation of Novel Approaches to
Software Engineering, Setubal, Portugal, 346-354.
https://doi.org/10.17993/3cemp.2022.110250.250-261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022
260
[15] Patel, A., and Jain, S. (2018). Formalisms of representing knowledge.
Procedia Computer
Science, 125, 542-549. doi:/10.1016/j.procs.2017.12.070.
[16] Ruttkay, Z., and Pelachaud, C. (Eds.). (2004).
From Brows to Trust: Evaluating Embodied
Conversational Agents. New York, Boston: Kluwer Academic Publishers.
[17] Simonova, A.L., Kapitonova, M.G. (2020). Frame model of knowledge representation in the
expert system of technological route selection.
Proceedings of International Scientific Technical
Conference on Innovative Engineering Technologies, Equipment and Materials, Conference
Series: Material Science and Engineering 915, Kazan, Russia, 1-4.
[18] Skachkova, N., Kruijff-Korbayová, I. (2021). Automatic Assignment of Semantic Frames in
Disaster Response Team Communication Dialogues.
Proceedings of the 14th International
Conference on Computational Semantics, Groningen, The Netherlands (Online), 93-109.
[19] Steels, L. (1978). Frame-Based Knowledge Representation
. USA: MIT Artificial Intelligence
Laboratory.
[20] Stickles, E., David, and O., Sweetser (2014). Grammatical constructions, frame structure, and
metonymy: Their contributions to metaphor computation. Proceedings of the 11th
High Desert
Linguistics Society Conference, Albuquerque, New Mexico City, 317-345.
[21] Torrent, T. T. et al. (2022). Representing Contexts in FrameNet: A Multidimensional, Multimodal
Approach. Frontiers in Psychology, (13), doi: 10.3389/fpsyg.2022.838441
[22] Thagard, P. (1984). Frames, knowledge, and Inference. Synthese, 61(2), 233-259. doi.org/
10.1007/BF00485316
[23] Thomson, B. (2013). Statistical Methods for Spoken Dialogue Management. London: Springer-
Verlag.
[24] Verdaguer, I. (2020). Semantic frames and semantic networks in the Health Science Corpus.
Estudios de
lingüística del español, (Anejo 1), 117-155.
https://doi.org/10.17993/3cemp.2022.110250.250-261
261
3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376
Ed. 50 Vol. 11 N.º 2 August - December 2022