logseq/pages/hls__SCICO-D-23-00080_R1_reviewer_1693666505321_0.md

file-path:: ../assets/SCICO-D-23-00080_R1_reviewer_1693666505321_0.pdf

• file:: SCICO-D-23-00080_R1_reviewer_1693666505321_0.pdf file-path:: ../assets/SCICO-D-23-00080_R1_reviewer_1693666505321_0.pdf
• Request for Proposals ls-type:: annotation hl-page:: 1 hl-color:: yellow id:: 650b029c-659a-49e5-936c-9d54df4ba6c4
• Request for Information ls-type:: annotation hl-page:: 1 hl-color:: yellow id:: 650b02a0-f225-4f6b-b888-f2b6c11c605c
• gap in leveraging model-based techniques in document-centric phases of proposal development, requirements analysis, etc ls-type:: annotation hl-page:: 1 hl-color:: green id:: 650b037e-d40f-4ae5-96c3-0b3ba3534647
• Large organizations respond to huge volumes of Request for Proposals (RFP)/ Request for Information (RFI) every year. The process of developing a proposal for an RFP/ RFI is completely manual and time-, effort-, and intellect-intensive. ls-type:: annotation hl-page:: 7 hl-color:: green id:: 650f0245-8835-49dd-886b-085f88e61243
• This paper presents an automated proposal development approach for a clientsupplied RFP/RFI using a combination of model-based and AI-enabled techniques and describes the case study of its successful deployment to hundreds of presales users across multiple geographies ls-type:: annotation hl-page:: 7 hl-color:: green id:: 650f0267-1449-430f-97f5-907787d8f388
• An RFP is a “collection of formal documents that includes a description of the desired form of response from a potential supplier, the relevant statement of work for the supplier, and required provisions in the supplier agreement” (ISO/IEC/IEEE 24765: 2017). ls-type:: annotation hl-page:: 7 hl-color:: purple id:: 651085c3-3715-404e-9100-5c24ebef1976
• An RFI typically precedes an RFP and seeks information about a product, service, or supplier capability. ls-type:: annotation hl-page:: 7 hl-color:: purple id:: 651085d8-2efb-46fc-91c5-aae0fef6af0c
• each business unit (BU) responds to hundreds of customer RFx yearly. ls-type:: annotation hl-page:: 7 hl-color:: blue id:: 65108639-4906-446c-83bc-e145bb84cc15
• An RFx has approximately 30 to 50 questions, and interpreting all these questions and developing a proposal by referring to various available sources of information is a time-consuming manual activity. ls-type:: annotation hl-page:: 7 hl-color:: purple id:: 65108659-4b6e-4c10-b9ca-8f120a4bfc9e
• Manual search is carried out by searching for specific documents in local/ shared repositories, searching for keywords, selecting relevant documents, and marking relevant portions of reusable tex ls-type:: annotation hl-page:: 7 hl-color:: blue id:: 6510867c-9f0d-4e69-ba8a-191cf2321d93
• no single source of truth and a lack of periodic knowledge validation resulting in inconsistent information and facts presented to customers ls-type:: annotation hl-page:: 7 hl-color:: green id:: 65108685-8435-4b2e-b8bb-1953eba43623
• Companies bid for multiple RFx and a lot of standardized content is reused across the proposals ls-type:: annotation hl-page:: 8 hl-color:: green id:: 65108947-ab2c-447d-a1ff-9cebfb80faa6
• A Day in the life of Bid ls-type:: annotation hl-page:: 8 hl-color:: yellow id:: 65108970-9dce-43c0-aa87-60270a0002af
• These documents are maintained in local repositories in word/ excel formats. ls-type:: annotation hl-page:: 8 hl-color:: yellow id:: 651089e7-5fa9-44e7-a613-57e128bd7a56
• AI and knowledge-aware software have remained largely untouched by MDE, and it has the potential to drastically change the way MDE is used and perceived in the software communit ls-type:: annotation hl-page:: 8 hl-color:: green id:: 65108a16-94e2-4397-8a2f-678e48d87e2d
• Here we propose an approach for automating the proposal development by exploring model-based techniques in combination with AI techniques. ls-type:: annotation hl-page:: 8 hl-color:: purple id:: 65108a23-1148-47c8-8277-20e98018b7bd
• uper-structure ls-type:: annotation hl-page:: 9 hl-color:: yellow id:: 65108a8e-fcee-4d96-b310-5f4801a6f54a
• This is a model-based AI intervention to generate a proposal in the desired format for a client-provided RFP/RFI. At the heart of the system is the “Proposal System Model” that enables a super-structure to be imposed on the information of interest lying around in a fragmented form as NL -text. This purpose-specific meta-model serves as the lens to extract information from the NL documents and facilitate document-to-model transformation in which the NL documents are parsed, and knowledge information is extracted into a machine-processable proposal knowledge model. Once a new client-provided RFP/RFI document is loaded, instantaneously on-click, leveraging the proposal knowledge model populated, solution recommendation can be generated for all RFx questions that have some relevant information in the knowledge model. As the complete information is in the model form, using model-to-text transformation, the proposal document can be automatically generated based on choices on recommendations. ls-type:: annotation hl-page:: 9 hl-color:: yellow id:: 65108acf-c3a9-421a-b8ad-9112ec9489b0
• document-to-models transformation for automatic parsing of knowledge documents into machineprocessable models ls-type:: annotation hl-page:: 9 hl-color:: green id:: 65108b1d-6d99-40c8-ab82-343f20cbb123
• knowledge-map recommendations for RFx questions ls-type:: annotation hl-page:: 9 hl-color:: green id:: 65108b2c-2b81-4d84-bd02-d1a945d18c57
• digitalized Proposal system by harmonizing multiple realization technologies ls-type:: annotation hl-page:: 9 hl-color:: green id:: 65108e98-8b87-42d5-8fac-16ee5d4c9c0d
• roposal document ls-type:: annotation hl-page:: 9 hl-color:: yellow id:: 65108eb3-048e-470a-9f8b-1dc53786e00f hl-stamp:: 1695583925101
• proposed evaluation metrics of interest? In general the evaluation is not presented in a systematic manner and it does not provide information on the evaluation set up, research questions, et ls-type:: annotation hl-page:: 6 hl-color:: yellow id:: 65108f57-1e68-4490-8ea4-4de3c4972929
• In the evaluation a quote from one of the users says"The RFx technology automates approximately 70% 75% " Which parts of the RFx development are not automated? What is the quality of the generated solutions? ls-type:: annotation hl-page:: 6 hl-color:: yellow id:: 65108f66-78b3-45bb-b76d-a9d41d36884a
• Proposal system model: A domain-agnostic meta-model for proposal development for any client-provided RFP/RFI is defined. ls-type:: annotation hl-page:: 10 hl-color:: green id:: 65108f73-7d06-4be7-866f-881f06763107
• DocToModels ls-type:: annotation hl-page:: 10 hl-color:: green id:: 65108f82-5d4a-4709-bcf4-1167f01a40c7
• it is used to populate knowledge data into the Proposal model. ls-type:: annotation hl-page:: 10 hl-color:: green id:: 65108fc5-f8c5-40b7-858e-cfc575fd962b
• K-Map recommende ls-type:: annotation hl-page:: 10 hl-color:: green id:: 65108fca-3253-4bda-95f6-25d0cde25396
• ModelToText ls-type:: annotation hl-page:: 10 hl-color:: green id:: 65108fd8-8f38-472f-8085-98eb03074d80
• unique harmonization of multiple technologies ls-type:: annotation hl-page:: 10 hl-color:: yellow id:: 65108feb-8432-4876-ba35-3d1b9d57e49e
• Lucene-based search engine facilitates knowledge search ls-type:: annotation hl-page:: 10 hl-color:: yellow id:: 65109043-3284-4742-9433-d7967d3ebc4e hl-stamp:: 1695584325195
• This Proposal_MM can be seen as a combination of proposal knowledge and solution model. ls-type:: annotation hl-page:: 11 hl-color:: green id:: 65109098-d5a0-4429-8924-65c128a9d564
• with data specific to the RFx to be responded to. ls-type:: annotation hl-page:: 11 hl-color:: yellow id:: 6510a332-3685-46b1-a58b-a4bb49e0d8ff
• banking, insurance, financial services, business process services, and so on ls-type:: annotation hl-page:: 11 hl-color:: green id:: 6510a35e-5860-4bda-a116-731e8500e8fa
• A KnowledgeType can have multiple KTypes. For the proposal process, solution knowledge for an offering can be classified into KnowledgeTypes such as product knowledge, process knowledge, facts knowledge, case knowledge and so on ls-type:: annotation hl-page:: 11 hl-color:: green id:: 6510a381-f8d0-4354-971a-3491a38e428e
• standard document structure in terms of organization of headings, styles, ls-type:: annotation hl-page:: 13 hl-color:: yellow id:: 6510a3aa-5ac2-4adf-acb8-01a1aa695eb0
• Using the meta-model as a lens, the document structural elements such as headings, paras, bullets, tables, etc. are analysed as per pattern mapping and corresponding model elements are automatically instantiated ls-type:: annotation hl-page:: 13 hl-color:: green id:: 6510a3be-21e4-4bf4-8c6b-21a650c0a0ee
• Pattern mapping is prepared based on the document structure and information organization of the document template used for capturing the solution knowledge. ls-type:: annotation hl-page:: 13 hl-color:: green id:: 6510a3df-c453-48f7-bbc5-6ad6dddedc77
• This content may have images, styled text, and tables. ls-type:: annotation hl-page:: 13 hl-color:: green id:: 6510a435-94de-4e12-a9a4-f691e4beef09
• ModelToText generates a proposal document as per desired document format using the reviewed K-Map. The KMap recommendations are composed as per compositional rules using model-to-text transformation ls-type:: annotation hl-page:: 15 hl-color:: yellow id:: 6510a48e-5aad-4067-8fac-f0a6449b742e
• The Proposal generation system depends largely on the proposal knowledge model. ls-type:: annotation hl-page:: 15 hl-color:: green id:: 6510a4f5-e26a-4e2c-9ee9-1ace88fa5f23
• A learning-based approach depicted in Fig. 7 is proposed to mine the relevant knowledge from submitted proposal documents ls-type:: annotation hl-page:: 15 hl-color:: green id:: 6510a53f-03c4-4f28-81e6-38576362c58b
• As proposal documents may have different document structures ls-type:: annotation hl-page:: 15 hl-color:: yellow id:: 6510a554-e14d-4ec9-a62a-ea44c185aa75
• a configurable extraction pattern-based approach is used for programmatically extracting question text and answer text. As per a standard pattern configuration, the question is extracted from the document heading and style information. The text after the question until the next question or the end of the document is considered an answer. ls-type:: annotation hl-page:: 15 hl-color:: yellow id:: 6510a5df-3cd0-43b1-a38b-71343422bf55
• he Maxent algorithm assumes the conditional dependence of the features. The extracted proposal questions are then classified by Question Classifier into concerns, and concern-specific knowledge data is generated. Automated checks are carried out on whether the question matches the existing Knowledge Element question. If so, the answer is checked for similarity, and the answer is marked for update. If it is a new question, then a new KnowledgeElement is proposed to be added with association to the appropriate concern. Human is in the loop to review/ refine these knowledge upgrade recommendations. This reviewed data is used for incremental knowledge upgrades to the proposal knowledge model.1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465 ls-type:: annotation hl-page:: 15 hl-color:: yellow id:: 6510a626-100c-4ef1-9fba-0b61bcd54da0
• The Maxent algorithm assumes the conditional dependence of the features. The extracted proposal questions are then classified by Question Classifier into concerns, and concern-specific knowledge data is generated. ls-type:: annotation hl-page:: 15 hl-color:: yellow id:: 6510a646-603b-439c-96ea-a77533810599 hl-stamp:: 1695589960697
• Initial knowledge documents are manually created by collating information from diverse sources. It is required to perform incremental knowledge updates in a periodic manner ls-type:: annotation hl-page:: 15 hl-color:: blue id:: 65117186-079b-4295-b4e2-4055a364a186
• sales and pre-sales teams (proposal writing experts) of a large-scale Business Unit (BU) that offers a suite of IT-enabled services to multiple customers across various geographies. ls-type:: annotation hl-page:: 16 hl-color:: green id:: 65117266-7041-44f0-8b7f-2b0cf29b3ad6
• Thousands of Requests for Proposal (RFP) / Requests for Information (RFI) floated by customers are responded to every year by that BU ls-type:: annotation hl-page:: 16 hl-color:: green id:: 65117274-1718-4b1b-bfc1-b86dfd62f91f
• It was a heavily expert-dependent process to select the solution fragments from documents relevant to a specific question in the RFx. ls-type:: annotation hl-page:: 16 hl-color:: green id:: 65117495-a5c1-415b-84e2-3fc9113e652b
• Manual search, copy-paste of content from individual documents, and formatting for appropriate styles was time-consuming and effort-intensive activit ls-type:: annotation hl-page:: 16 hl-color:: green id:: 651174a4-2d25-43f6-bd37-6ad3b37c825b
• 700 BU users from sales/ pre-sales teams ls-type:: annotation hl-page:: 16 hl-color:: green id:: 651174ad-e656-4e13-b0cc-9b59fb1eb163
• 3.1. Domain and Offerings ls-type:: annotation hl-page:: 16 hl-color:: yellow id:: 6511750e-a50b-40d0-98ba-1ad38c3ceb0f
• The Proposal system has been deployed and rolled out to over 700 BU users from sales/ pre-sales teams. Solution documents related to the capabilities of 31 offerings were collated and parsed using the DocToModels. ls-type:: annotation hl-page:: 16 hl-color:: yellow id:: 65117555-0030-4899-9f5e-5e6704321216
• Fig. 8 Knowledge Model View ls-type:: annotation hl-page:: 17 hl-color:: yellow id:: 651175a7-aece-49f3-b6e9-ad473abae0ee
• Each Question in the RFP is analyzed and transformed into a query to retrieve relevant proposal Knowledge Elements for that question as a K-Map. ls-type:: annotation hl-page:: 18 hl-color:: yellow id:: 6511769b-73ae-4190-8878-a335a888893c
• Hundreds of proposals have been generated across seven geographies using the Proposal system. ls-type:: annotation hl-page:: 20 hl-color:: blue id:: 65117731-fe4f-4e22-ba61-3e70c66f8ed1
• ese 15 cases represent a wide range of RFx scenarios across domains and geographies ls-type:: annotation hl-page:: 20 hl-color:: blue id:: 65117766-fbe8-42fd-9d49-b3cd5307317a
• RFx and RFx Context are populated, and on search on any RFx, solution recommendations are generated question-wise ls-type:: annotation hl-page:: 20 hl-color:: green id:: 65117b7d-e694-4276-8a0f-560edb751095
• automates approximately 70% - 75% effort in preparation of the RFP/RI responses ls-type:: annotation hl-page:: 23 hl-color:: green id:: 65117bd4-4128-4a69-bba5-41a1c4e23889
• expert-driven proposal development for RFP/ RFIs, a model-based and AI-enabled approach is proposed to help the pre-sales teams in automatically generating a standard proposal document and provide a kick-start to the proposal process ls-type:: annotation hl-page:: 23 hl-color:: purple id:: 65117bfa-52f9-44fd-8933-ef4941495454 hl-stamp:: 1695733490005
• abstracts the critical elements from diverse knowledge documents belonging to various domains and offerings and the key constructs of a proposal document ls-type:: annotation hl-page:: 23 hl-color:: green id:: 65117c10-8703-4235-aaf3-cedc4ff70901
• The Model-Driven Engineering (MDE) based approach offered a plethora of advantages for the proposal system, significantly improving its development and maintenance process. ls-type:: annotation hl-page:: 23 hl-color:: yellow id:: 65117c30-8209-4797-a457-ff9e5bfe1ca9
• This capability to generate a “proposal document” is a big value-add to the presales teams in the industry. ls-type:: annotation hl-page:: 23 hl-color:: green id:: 6512e010-966f-4245-9e9a-5a01f95c1676
• We could easily modify the models to reflect new information. ls-type:: annotation hl-page:: 23 hl-color:: green id:: 6512e04a-23e1-4982-bfb1-dd8a505e5b55
• y leveraging existing modeling infrastructure for populating, querying we could rapidly create the proposal system ls-type:: annotation hl-page:: 23 hl-color:: yellow id:: 6512e053-caf4-4086-aa3c-333adb8e98c6
• The initial hurdle was in understanding the diverse knowledge document formats received from presales teams. ls-type:: annotation hl-page:: 23 hl-color:: green id:: 6512e08d-dbc5-45db-8208-46d203804be0
• proposal system meta-model ls-type:: annotation hl-page:: 23 hl-color:: blue id:: 6512e09d-be01-4cb7-ac9b-ce7442926dbb hl-stamp:: 1695735972537
• domain, offering, concern, knowledge element, and so on. ls-type:: annotation hl-page:: 23 hl-color:: blue id:: 6512e0a2-20c9-4268-af1d-e42442a5399d hl-stamp:: 1695735974764
• The meta-model enables a super-structure to be imposed on the information of interest lying around in fragmented form as NL text and is the central building block in our architecture to which every other component feeds/ retrieves model data ls-type:: annotation hl-page:: 23 hl-color:: green id:: 6512e153-8ac2-48bf-a426-2907279191ae
• The next key challenge was handling the diversity in document structures and formats while parsing information from NL documents. The structure of the document provided insights into the information organization and could be mapped to the meta-model concepts ls-type:: annotation hl-page:: 23 hl-color:: yellow id:: 6512e165-d6e3-43b1-b085-a8e92e2c3555
• defining extraction patterns for mapping information in the document to the meta-model. ls-type:: annotation hl-page:: 23 hl-color:: yellow id:: 6512e184-4384-4e04-b59f-26e64d693569 hl-stamp:: 1695736200053
• This approach is both document structure agnostic and metamodel agnostic, thus paving the way forward to extend this technology for other business use cases that are manual and document-centric such as requirements. ls-type:: annotation hl-page:: 24 hl-color:: yellow id:: 6512e1da-5919-417f-a514-0227aec42646
• For question analysis, various matching heuristic such as exact matching, inexact matching, and fuzzy matching was implemented ls-type:: annotation hl-page:: 24 hl-color:: green id:: 6512e2b7-67a0-4201-a25d-7d75ac617fc9
• for an RFP from a UK Client, to respond to a question on providing references on past experience, the knowledge elements retrieved need to be past case studies from UK geography. ls-type:: annotation hl-page:: 24 hl-color:: yellow id:: 6512e3c3-bbe3-420c-adce-a1e072e89e8c hl-stamp:: 1695736774387
• Dynamic field population was another challenge faced during proposal document composition. ls-type:: annotation hl-page:: 24 hl-color:: green id:: 6512e3f6-ffca-471d-8ce0-785cc2df7f12
• There were many operational challenges during the deployment of this technology ls-type:: annotation hl-page:: 24 hl-color:: green id:: 6512e436-c7ed-4d0e-aba8-edb7a87fccef
• This proposal development is an expertise-driven, heavily manual activity, and is centered around various types of natural language documents prepared, referred, and maintained. ls-type:: annotation hl-page:: 25 hl-color:: green id:: 6512e496-0525-418f-895f-4d9d7727451c
• utomatic generation of formatted Proposal document with ~70% of questions answered ls-type:: annotation hl-page:: 25 hl-color:: purple id:: 6512e4b5-7321-48f3-ad01-7b857813ba6e