Article Preview
TopIntroduction
In today’s globalized economy, economic and trade relations between various countries have become closer, leading to a significant increase in the proportion of international bidding. To adapt to this process, companies are working to enhance the market competitiveness of international engineering contracting and trying to expand the international market for a larger share of business (Tso et al., 2019). In the international bidding process, the first and foremost thing to make a targeted bid is to read and understand the bidding documents (ITB, construction to bid). Differences in the business credit environment and the existence of qualification and honor systems in different countries make the bidding process more brutal during the competition of project proposals. The regulatory agencies in each country and region are equally strict about international bidding, so they are extra cautious in conducting an international bidding process (Markiewicz et al., 2022). Different companies often have significant differences in the bidding process owing to the form of document writing and different countries’ laws and regulations as well as language habits. These processes require huge human and material resources for relevant work analysis to complete the accurate understanding of the bidding documents. Consider engineering bidding documents as an example. The relevant information involved in the international bidding process is shown in Figure 1 (Ahmed et al., 2022).
Figure 1. The Structure of Bidding Documents in Construction
Figure 1 shows that the information covered by the bidding documents is extremely cumbersome, and in addition to the requirements for the main content, there are many requirements for the relevant content of the annexes. In addition to the requirements being high, the risks of inconsistencies in the bidding content, irregularities in the compilation form, and unclear restrictive clauses may occur owing to the mistakes of the document management department during the bidding (Liu et al., 2020). The existing management mode of the standard model bidding document is no longer able to cope with the demand of the bidding work, and therefore, it has become urgent to innovate a modular management mode of the model bidding document that is suitable for the development needs (Ghavidel et al., 2019). The main purpose is to unify the chapter structure and content form of all types of model bidding documents, assign the same part of content to curing modules according to different situations, form common use and reuse clauses, and replace different parts to form module management. The prerequisite for module management is the detailed structural division of the bidding documents and the feature screening of the relevant parts. With the development of artificial intelligence (AI), it has become possible to apply natural language processing (NLP) technology to the structural analysis of bidding documents (Kim et al., 2020).
Deep learning can accelerate the analysis process by extracting features from documents. In the field of text analysis, the separation of linguistic features by word separation techniques followed by sentiment analysis and topic extraction of word vector features using classical networks such as recurrent neural network (RNN) and long short-term memory (LSTM) is the current focus of NLP research in text analysis (Di et al., 2020). For the bidding documents, the key problem to be solved by AI is how to extract the unit information under each heading and provide the relevant staff with more fine-grained and precise knowledge based on the already decomposed documents. Therefore, this paper proposes a structured analysis method for two-layer bidding documents based on the characteristics and the needs of bidding documents. Combining this method with the advantages of AI in text analysis to achieve the task of document structure division and key information extraction, we include the following contributions in this paper: