PcMSP A Dataset for Scientiﬁc Action Graphs Extraction from Polycrystalline Materials Synthesis Procedure Text Xianjun Yang1 Ya Zhuo2 Julia Zuo2 Xinlu Zhang1 Stephen Wilson2 Linda Petzold1

2025-04-29 0 0 620.31KB 14 页 10玖币

侵权投诉

PcMSP: A Dataset for Scientiﬁc Action Graphs Extraction from

Polycrystalline Materials Synthesis Procedure Text

Xianjun Yang1, Ya Zhuo2, Julia Zuo2, Xinlu Zhang1, Stephen Wilson2, Linda Petzold1

1Department of Computer Science 2Department of Materials Science and Engineering

University of California, Santa Barbara

{xianjunyang, yzhuo, jlzuo, xinluzhang, stephendwilson, petzold}@ucsb.edu

Abstract

Scientiﬁc action graphs extraction from ma-

terials synthesis procedures is important for

reproducible research, machine automation,

and material prediction. But the lack of

annotated data has hindered progress in

this ﬁeld. We demonstrate an effort to

annotate Polycrystalline Materials Synthesis

Procedures (PcMSP) from 305 open access

scientiﬁc articles for the construction of syn-

thesis action graphs. This is a new dataset for

material science information extraction that si-

multaneously contains the synthesis sentences

extracted from the experimental paragraphs, as

well as the entity mentions and intra-sentence

relations. A two-step human annotation and

inter-annotator agreement study guarantee the

high quality of the PcMSP corpus. We intro-

duce four natural language processing tasks:

sentence classiﬁcation, named entity recogni-

tion, relation classiﬁcation, and joint extrac-

tion of entities and relations. Comprehen-

sive experiments validate the effectiveness of

several state-of-the-art models for these chal-

lenges while leaving large space for improve-

ment. We also perform the error analysis and

point out some unique challenges that require

further investigation. We will release our an-

notation scheme, the corpus, and codes to the

research community to alleviate the scarcity of

labeled data in this domain1.

1 Introduction

Synthesis procedural texts are written in instruc-

tional languages (Grishman,2001;Grishman and

Kittredge,2014) to represent the step-by-step reac-

tions, but also contain the distinct features in spe-

ciﬁc domains, such as the domain notations, writ-

ing styles, and journal requirements. The synthesis

procedures of materials science articles include

valuable information for new materials prediction

(Raccuglia et al.,2016), laboratory automation (Co-

ley et al.,2019) and knowledge graph construction

1https://github.com/Xianjun-Yang/PcMSP

Synthesis Paragraph

Polycrystalline[Descriptor]

sample of

composition

Sr2CoO4[Material_target]

was

synthesized[operation]

under

high pressure[Property_pressure]

high temperature[Property_temperature]

. Start-

ing materials of

SrO2[Material_recipe]

and

Co[Material_recipe]

were

well[Descriptor] mixed[operation]

in a

molar ratio[Descriptor]

SrO2[Material_recipe]

Co[Material_recipe]

2 : 1[Value]

. The

mixture[Material-intermedium]

was

sealed[operation]

into

a[Value] gold[Descriptor] capsule[Device]

. ...

The crystal structure of the polycrystalline sample

was identiﬁed by the powder X-ray diffraction

(XRD, Rigaku Smart- lab3), using Cu-K

radiation (λ=1.54184Å). ...

Table 1: An example of a synthesis paragraph from our

dataset with index srep27712 (Li et al.,2016).

(Mrdjenovich et al.,2020). However, available

datasets are extremely limited, despite the notable

work by (Mysore et al.,2017,2019;Friedrich et al.,

2020;O’Gorman et al.,2021).

The goal of information extraction from proce-

dures is to construct the action graphs, which refer

to all the steps in a synthesis making up a Directed

Acyclic Graph (DAG) (Mysore et al.,2019;Kulka-

rni et al.,2018) (as can be seen from one example

in Figure 1). This can be further breakdown into

three tasks: sentence classiﬁcation, named entity

recognition (NER), and relation extraction (RE).

Previous research (Mysore et al.,2017,2019) ei-

ther annotates the whole synthesis paragraph in

the general inorganic domain, ignoring the non-

synthesis sentences and subdomain discrepancy or

only focuses on entity mentions (Friedrich et al.,

2020;O’Gorman et al.,2021).

To ﬁll this gap, we focus on one important cat-

egory of polycrystalline materials and simultane-

arXiv:2210.12401v1 [cs.CL] 22 Oct 2022

Figure 1: A synthesis action graph constructed from

Table 1.

ously include all three tasks. The annotation guide-

lines are designed by materials experts after com-

prehensive discussion, and the new dataset is sub-

sequently labeled with a two-round annotation.

The key contributions of this paper include:

•

We contribute a new large-scale dataset, as

well as an annotation scheme with high qual-

ity for information extraction in materials sci-

ence.

•

We conduct comprehensive experiments on

four tasks, sentence classiﬁcation, named en-

tity recognition, relation extraction, and joint

extraction to provide baselines.

•

We perform error analysis and point out

unique challenges and potential use of this

dataset for future research.

2 Related Work

Scientiﬁc information extraction

With the fast-growing volume of scholarly pub-

lications, it is highly demanding to extract struc-

tured information from large-scale scientiﬁc liter-

ature in many domains (Augenstein et al.,2017;

Luan et al.,2018;Jiang et al.,2019;Beltagy et al.,

2019;Buscaldi et al.,2019), like biomedical do-

main (Shah et al.,2003;Lai et al.,2021;Zhang

et al.,2021;Lewis et al.,2020;Kulkarni et al.,

2018) and chemistry domain (Rocktäschel et al.,

2012;He et al.,2020). In the ﬁeld of materials sci-

ence, there have been few attempts in this direction,

leaving many unexplored challenges for research

(Hong et al.,2021). Recent research mainly fo-

cuses on knowledge base construction (Jiang et al.,

2019;Luan et al.,2018), new materials discovery

(Isayev,2019), and automation of lab procedures

(Vaucher et al.,2020;Tamari et al.,2021;Steiner

et al.,2019). (Beltagy et al.,2019) trained a Bidi-

rectional Encoder Representations from Transform-

ers model (SciBERT) on 1.14M scientiﬁc papers

from Semantic Scholar for scientiﬁc information

extraction.

Materials procedures information extraction

In the area of annotation of materials synthesis

procedures, (Mysore et al.,2019) annotate 230

general materials synthesis paragraphs for NER

and RE tasks. Similar work is also undertaken

by (Friedrich et al.,2020), in which 45 open ac-

cess scholarly articles are labeled for experiment-

describing sentence classiﬁcation, NER, and slot

ﬁlling tasks. However, in contrast to our works,

their annotation scheme focuses on the full text

rather than the experimental section. (Kuniyoshi

et al.,2020) annotate the synthesis process of all-

solid-state batteries from the scientiﬁc literature,

but their corpus is not publicly available. (Walker

et al.,2021) release MatBERT trained on 50 million

materials science paragraphs to explore the impact

of domain-speciﬁc pre-training on NER task. Also

of interest, (O’Gorman et al.,2021) recently create

the largest corpus for entity mentions extraction in

both general domain and subdomain from material

synthesis text, but the relations between entities are

still missing.

Named entity recognition and relation

extraction

Many neural network-based models have been pro-

posed for named entity recognition, for example,

(Huang et al.,2015;Lample et al.,2016;Panchen-

drarajan and Amaresan,2018). The core idea uses

one encoding layer (e.g. Long Short-Term Mem-

ory (LSTM) (Hochreiter and Schmidhuber,1997),

BERT) for representation and one additional condi-

tional random ﬁelds (CRF (Lafferty et al.,2001))

layer for sequence labeling. Then relations are pre-

dicted based on either gold entities or predicted

entities, and PURE (Zhong and Chen,2021) de-

signs two separate encoders for joint extraction of

Figure 2: An annotated PcMSP example on the INCEpTION platform, taken from srep15507 (Man et al.,2015).

entities and relations. We adopt their model for our

tasks due to its super performance.

3 The Selection of Our Dataset

Here we talk about the importance of our selec-

tion and how is it different from other materials

procedural text corpora.

Why do we choose inorganic polycrystalline

materials?

There are a number of sub-categories

within solid-state inorganic materials. For exam-

ple, materials can be divided based on function and

properties, such as the battery or thermoelectric

materials. Synthesis within both categories largely

falls within the broader category of solid-state syn-

thesis and even then, there is a high degree of over-

lap with other function categories, such as quantum

and magnetic materials. More importantly,

those

materials are usually in the form of polycrys-

talline

. Other subcategories relate to form factors,

for instance, single-crystalline synthesis often starts

with a

polycrystalline

synthesis and therefore has

a high degree of overlap with solid-state synthesis.

Inorganic polycrystal compounds span combina-

tions of the entire periodic table and different chem-

ical bonding schemes, such that their synthesis typi-

cally takes place under extreme conditions, such as

high temperature and pressure. Reaction pathways

are therefore difﬁcult to characterize without spe-

cialized equipment and are not well established for

any given material. In particular, solid-state reac-

tions, which are the main techniques to synthesize

inorganic polycrystalline materials, are particularly

similar to a “black box”, where materials scientists

can only make educated guesses to the procedure

or stability of a new reaction. This presents a prime

opportunity (Mysore et al.,2017,2019) for compil-

ing published inorganic synthesis data in order to

demystify the black box of solid-state inorganic ma-

terials synthesis and create datasets for future text

mining endeavors. While there have been efforts

within general solid-state materials (Mysore et al.,

2017,2019;O’Gorman et al.,2021) and battery

materials subcategory (Friedrich et al.,2020), this

work aims to extend the subcategory of inorganic

solid-state synthesis methods in order to address

the frequent overlap and “borrowing” of materials

between subdisciplines of materials science.

Why do we discard characterization sen-

tences?

Inorganic reactions typically involve rel-

atively few reactions from a set of precursors and

there are very few puriﬁcation pathways for solid

materials compared to organic materials or liq-

uids. Therefore, characterizations of solid-state

inorganic reactions are seldom reported in litera-

ture unless they proceed to complete purity within

standard measurement ﬁdelity. This is in contrast

to organic materials where there are a number of

important characterization metrics in a compound,

such as molecular weight in polymers or reaction

yield. Therefore, these standard characterization

measurements do not add valuable information for

a researcher attempting to recreate the reported

synthesis method and we decide to discard these

characterization sentences.

Why do we annotate sentence, entity, and re-

lation simultaneously?

A full action graph con-

sists of both entities and relations extracted from

experimental-describing sentences. However, most

previous research either ignores the annotation of

sentence or relation information, making them in-

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

PcMSP:ADatasetforScienticActionGraphsExtractionfromPolycrystallineMaterialsSynthesisProcedureTextXianjunYang1,YaZhuo2,JuliaZuo2,XinluZhang1,StephenWilson2,LindaPetzold11DepartmentofComputerScience2DepartmentofMaterialsScienceandEngineeringUniversityofCalifornia,SantaBarbara{xianjunyang,yzhuo,jlzuo,...

展开>> 收起<<

PcMSP A Dataset for Scientiﬁc Action Graphs Extraction from Polycrystalline Materials Synthesis Procedure Text Xianjun Yang1 Ya Zhuo2 Julia Zuo2 Xinlu Zhang1 Stephen Wilson2 Linda Petzold1.pdf

共14页,预览3页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

PcMSP A Dataset for Scientiﬁc Action Graphs Extraction from Polycrystalline Materials Synthesis Procedure Text Xianjun Yang1 Ya Zhuo2 Julia Zuo2 Xinlu Zhang1 Stephen Wilson2 Linda Petzold1

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: