Structured chemical reaction information plays a vital role for chemists
engaged in laboratory work and advanced endeavors such as computer-aided drug
design. Despite the importance of extracting structured reactions from
scientific literature, data annotation for this purpose is cost-prohibitive due
to the significant labor required from domain experts. Consequently, the
scarcity of sufficient training data poses an obstacle to the progress of
related models in this domain. In this paper, we propose ReactIE, which
combines two weakly supervised approaches for pre-training. Our method utilizes
frequent patterns within the text as linguistic cues to identify specific
characteristics of chemical reactions. Additionally, we adopt synthetic data
from patent records as distant supervision to incorporate domain knowledge into
the model. Experiments demonstrate that ReactIE achieves substantial
improvements and outperforms all existing baselines.

本篇研究文章提出了一种名为 ReactIE 的方法，通过弱监督预训练的方式，结合来自文本中的频繁模式和专利记录中的数据，以识别化学反应的特定特征。实验证明，ReactIE 优于所有现有的基准模型。

ReactIE：利用弱监督增强化学反应提取

ReactIE: Enhancing Chemical Reaction Extraction with Weak Supervision

Invention has been commonly conceptualized as a search over a space of
combinatorial possibilities. Despite the existence of a rich literature,
spanning a variety of disciplines, elaborating on the recombinant nature of
invention, we lack a formal and quantitative characterization of the
combinatorial process underpinning inventive activity. Here we utilize U.S.
patent records dating from 1790 to 2010 to formally characterize the invention
as a combinatorial process. To do this we treat patented inventions as carriers
of technologies and avail ourselves of the elaborate system of technology codes
used by the U.S. Patent Office to classify the technologies responsible for an
invention's novelty. We find that the combinatorial inventive process exhibits
an invariant rate of "exploitation" (refinements of existing combinations of
technologies) and "exploration" (the development of new technological
combinations). This combinatorial dynamic contrasts sharply with the creation
of new technological capabilities -- the building blocks to be combined --
which has significantly slowed down. We also find that notwithstanding the very
reduced rate at which new technologies are introduced, the generation of novel
technological combinations engenders a practically infinite space of
technological configurations.

本研究利用美国专利记录对 1790 年至 2010 年间的发明进行了形式化和量化的特征化，发现创新是一种组合过程，呈现出 “开发” 和 “探索” 的不变比率，而新技术产生的速度显著减缓，但产生的技术组合却多种多样。